netapp

Data ONTAP® 7.3Active/Active Configuration Guide

NetApp, Inc.495 East Java DriveSunnyvale, CA 94089 U.S.A.Telephone: +1 (408) 822-6000Fax: +1 (408) 822-4501Support telephone: +1 (888) 4-NETAPPDocumentation comments: [email protected] Web: www.netapp.com

Part number: 210-05044_B0Updated for Data ONTAP 7.3.5.1 on 11 May 2011

Contents

Active/active configuration types and requirements ................................. 9Overview of active/active configurations ................................................................... 9

What an active/active configuration is ............................................................ 9

Nondisruptive operations and fault tolerance with active/active

configurations ............................................................................................ 9

Characteristics of nodes in an active/active configuration ............................ 10

Systems with variable HA configurations ..................................................... 10

Best practices for active/active configurations .............................................. 13

Comparison of active/active configuration types .......................................... 14

Standard active/active configurations ....................................................................... 15

How Data ONTAP works with standard active/active configurations ......... 15

Standard active/active configuration diagram ............................................... 16

Setup requirements and restrictions for standard active/active

configurations .......................................................................................... 17

Configuration variations for standard active/active configurations .............. 18

Multipath HA requirements and recommendations ...................................... 18

Understanding mirrored active/active configurations ............................................... 21

Advantages of mirrored active/active configurations ................................... 22

Setup requirements and restrictions for mirrored active/active

configurations .......................................................................................... 22

Configuration variations for mirrored active/active configurations .............. 23

Understanding stretch MetroClusters ........................................................................ 23

Continued data service after loss of one node with MetroCluster ................ 24

Advantages of stretch MetroCluster configurations ...................................... 24

Stretch MetroCluster configuration with FAS systems ................................. 25

Stretch MetroCluster configuration on single enclosure active/active

configurations .......................................................................................... 26

How Data ONTAP works with stretch MetroCluster configurations ........... 26

Setup requirements and restrictions for stretch MetroCluster

configurations .......................................................................................... 27

Configuration variations for stretch MetroCluster configurations ................ 28

Understanding fabric-attached MetroClusters .......................................................... 28

Table of Contents | 3

Fabric-attached MetroClusters use Brocade Fibre Channel switches ........... 28

Advantages of fabric-attached MetroCluster configurations ........................ 29

Fabric-attached MetroCluster configuration for FAS systems ..................... 29

Fabric-attached MetroCluster configuration on single-enclosure active/

active configurations ............................................................................... 30

How Data ONTAP works with fabric-attached MetroCluster

configurations .......................................................................................... 31

Setup requirements and restrictions for fabric-attached MetroClusters ........ 31

Configuration limitations for fabric-attached MetroClusters ........................ 33

Configuration variations for fabric-attached MetroClusters ......................... 34

Active/active configuration installation .................................................... 35Systems with two controller modules in the same chassis ........................................ 35

System cabinet or equipment rack installation .......................................................... 36

Active/active configurations in an equipment rack ....................................... 36

Active/active configurations in a system cabinet .......................................... 36

Required documentation, tools, and equipment ........................................................ 36

Required documentation ............................................................................... 36

Required tools ............................................................................................... 38

Required equipment ...................................................................................... 38

Preparing your equipment ......................................................................................... 39

Installing the nodes in equipment racks ........................................................ 39

Installing the nodes in a system cabinet ........................................................ 40

Cabling a standard active/active configuration ......................................................... 40

Determining which Fibre Channel ports to use for Fibre Channel disk

shelf connections ..................................................................................... 41

Cabling Node A to DS14mk2 or DS14mk4 FC disk shelves ....................... 42

Cabling Node B to DS14mk2 or DS14mk4 FC disk shelves ........................ 44

Cabling the cluster interconnect (all systems except 32xx) .......................... 46

Cabling the cluster interconnect (32xx systems in separate chassis) ............ 47

Cabling a mirrored active/active configuration ......................................................... 47

Determining which Fibre Channel ports to use for Fibre Channel disk

shelf connections ..................................................................................... 48

Creating your port list for mirrored active/active configurations ................. 49

Cabling the channel A DS14mk2 or DS14mk4 FC disk shelf loops ............ 50

Cabling the channel B DS14mk2 or DS14mk4 FC disk shelf loops ............ 52

Cabling the redundant multipath HA connection for each loop .................... 54

4 | Data ONTAP 7.3 Active/Active Configuration Guide

Cabling the cluster interconnect (all systems except 32xx) .......................... 56

Cabling the cluster interconnect (32xx systems in separate chassis) ............ 57

Required connections for using uninterruptible power supplies with standard or

mirrored active/active configurations .................................................................. 57

MetroCluster installation ........................................................................... 59Required documentation, tools, and equipment ........................................................ 59

Required documentation ............................................................................... 59

Required tools ............................................................................................... 60

Required equipment ...................................................................................... 61

MetroCluster and software-based disk ownership .................................................... 63

Converting an active/active configuration to a fabric-attached MetroCluster .......... 63

Upgrading an existing MetroCluster ......................................................................... 65

Cabling a stretch MetroCluster ................................................................................. 67

Cabling a stretch MetroCluster between single-enclosure active/active

configuration systems .......................................................................................... 67

Changing the default configuration speed of a stretch MetroCluster ........................ 68

Resetting a stretch MetroCluster configuration to the default speed ........................ 70

Cabling a fabric-attached MetroCluster .................................................................... 71

Planning the fabric-attached MetroCluster installation ................................. 73

Configuration differences for fabric-attached MetroClusters on single-

enclosure active/active configuration ...................................................... 74

Configuring the switches ............................................................................... 75

Cabling Node A ............................................................................................. 76

Cabling Node B ............................................................................................. 82

Assigning disk pools (if you have software-based disk ownership) ............. 87

Verifying disk paths ...................................................................................... 88

Required connections for using uninterruptible power supplies with

MetroCluster configurations ................................................................................ 88

Reconfiguring an active/active configuration into two stand-alonesystems .................................................................................................... 91

Ensuring uniform disk ownership within disk shelves and loops in the system ....... 91

Disabling the active/active software ......................................................................... 92

Reconfiguring nodes using disk shelves for stand-alone operation .......................... 93

Requirements when changing a node using array LUNs to stand-alone ................... 95

Reconfiguring nodes using array LUNs for stand-alone operation ........................... 96

Configuring an active/active configuration .............................................. 99


Bringing up the active/active configuration .............................................................. 99

Considerations for active/active configuration setup .................................... 99

Configuring shared interfaces with setup .................................................... 100

Configuring dedicated interfaces with setup ............................................... 101

Configuring standby interfaces with setup .................................................. 101

Enabling licenses ..................................................................................................... 102

Setting options and parameters ............................................................................... 103

Option types for active/active configurations ............................................. 103

Setting matching node options .................................................................... 103

Parameters that must be the same on each node ......................................... 104

Disabling the change_fsid option in MetroCluster configurations ............. 104

Verifying and setting the HA state of a32xx controller module and

chassis .................................................................................................... 105

Configuring hardware-assisted takeover ..................................................... 106

Configuration of network interfaces ....................................................................... 109

What the networking interfaces do .............................................................. 109

IPv6 considerations in an active/active configuration ................................. 109

Configuring network interfaces for active/active configurations ................ 110

Configuring partner addresses on different subnets (MetroClusters only) . 115

Downloading and running the HA Configuration Checker utility .......................... 119

Testing takeover and giveback ................................................................................ 119

Managing takeover and giveback ........................................................... 121How takeover and giveback work ........................................................................... 121

When takeovers occur ................................................................................. 121

What happens during takeover .................................................................... 122

What happens after takeover ....................................................................... 122

What happens during giveback ................................................................... 122

Management of an active/active configuration in normal mode ............................. 123

Monitoring active/active configuration status ............................................. 123

Monitoring the hardware-assisted takeover feature .................................... 123

Description of active/active configuration status messages ........................ 126

Displaying the partner's name ..................................................................... 127

Displaying disk and array LUN information on an active/active

configuration .......................................................................................... 127

Enabling and disabling takeover ................................................................. 128

Enabling and disabling automatic takeover of a panicked partner .............. 128


Halting a node without takeover ................................................................. 128

Configuring automatic takeover .............................................................................. 129

Reasons for automatic takeover .................................................................. 129

Commands for performing a manual takeover ............................................ 131

Specifying the time period before takeover ................................................ 132

How disk shelf comparison takeover works ............................................... 133

Configuring VIFs or interfaces for automatic takeover .............................. 133

Takeover of vFiler units and the vFiler unit limit ....................................... 133

Managing an active/active configuration in takeover mode ................................... 134

Determining why takeover occurred ........................................................... 134

Statistics in takeover mode .......................................................................... 134

Managing emulated nodes ....................................................................................... 135

Management exceptions for emulated nodes .............................................. 135

Accessing the emulated node from the takeover node ................................ 135

Assessing the emulated node remotely ....................................................... 136

Emulated node command exceptions .......................................................... 137

Performing dumps and restores for a failed node ................................................... 139

Giveback operations ................................................................................................ 140

Performing a manual giveback .................................................................... 140

Configuring giveback .................................................................................. 142

Configuring automatic giveback ................................................................. 144

Troubleshooting takeover or giveback failures ....................................................... 144

Managing DS14mk2 or DS14mk4 FC disk shelves in an active/activeconfiguration ........................................................................................ 147

Adding DS14mk2 or DS14mk4 FC disk shelves to a multipath HA loop .............. 147

Upgrading or replacing modules in an active/active configuration ........................ 149

About the disk shelf modules ...................................................................... 149

Restrictions for changing module types ...................................................... 149

Best practices for changing module types ................................................... 150

Testing the modules .................................................................................... 150

Understanding redundant pathing in active/active configurations .............. 150

Determining path status for your active/active configuration ..................... 151

Hot-swapping a module .............................................................................. 153

Disaster recovery using MetroCluster .................................................... 155Conditions that constitute a disaster ........................................................................ 155

Ways to determine whether a disaster occurred .......................................... 155


Failures that do not require disaster recovery ............................................. 155

Recovering from a disaster ...................................................................................... 156

Restricting access to the disaster site node .................................................. 157

Forcing a node into takeover mode ............................................................. 158

Remounting volumes of the failed node ..................................................... 158

Recovering LUNs of the failed node ........................................................... 159

Fixing failures caused by the disaster ......................................................... 160

Reestablishing the MetroCluster configuration ........................................... 161

Nondisruptive operations with active/active configurations ................ 167Controller failover and single-points-of-failure ..................................... 169

Single-point-of-failure definition ............................................................................ 169

SPOF analysis for active/active configurations ...................................................... 170

Failover event cause-and-effect table ...................................................................... 173

Feature update record .............................................................................. 181Abbreviations ............................................................................................ 187Glossary ..................................................................................................... 203Copyright information ............................................................................. 205Trademark information ........................................................................... 207How to send your comments .................................................................... 209Index ........................................................................................................... 211


Active/active configuration types andrequirements

There are four types of active/active configurations, each having different advantages andrequirements.

Overview of active/active configurationsThe different types of active/active configurations all offer access to storage through two differentcontrollers. Each type has its own benefits and requirements.

What an active/active configuration isAn active/active configuration is two storage systems (nodes) whose controllers are connected toeach other either directly or, in the case of a fabric-attached MetroCluster, through switches and FC-VI interconnect adapters.

You can configure the active/active configuration so that each node in the pair shares access to acommon set of storage, subnets, and tape drives, or each node can own its own distinct set of storageand subnets.

Depending on the model, nodes are connected to each other through an NVRAM adapter, gigabitEthernet connections, or, in the case of systems with two controllers in a single chassis, through aninternal interconnect. This allows one node to serve data that resides on the disks of its failed partnernode. Each node continually monitors its partner, mirroring the data for each other’s nonvolatilememory (NVRAM or NVMEM).

Nondisruptive operations and fault tolerance with active/activeconfigurations

Active/active configurations provide fault tolerance and the ability to perform nondisruptiveupgrades and maintenance.

Configuring storage systems in an active/active configuration provides the following benefits:

• Fault toleranceWhen one node fails or becomes impaired a takeover occurs, and the partner node continues toserve the failed node’s data.

• Nondisruptive software upgradesWhen you halt one node and allow takeover, the partner node continues to serve data for thehalted node while you upgrade the node you halted.

• Nondisruptive hardware maintenanceWhen you halt one node and allow takeover, the partner node continues to serve data for thehalted node while you replace or repair hardware in the node you halted.

Active/active configuration types and requirements | 9

Related concepts

Nondisruptive operations with active/active configurations on page 167

Managing DS14mk2 or DS14mk4 FC disk shelves in an active/active configuration on page 147

Characteristics of nodes in an active/active configurationTo configure and manage nodes in an active/active configuration, you should be familiar with thecharacteristics of active/active configurations.

• The controllers in the active/active configuration are connected to each other either through acluster interconnect consisting of adapters and cable, or, in systems with two controllers in thesame chassis, through an internal interconnect. The nodes use the interconnect to do the followingtasks:

• Continually check whether the other node is functioning• Mirror log data for each other’s NVRAM or NVMEM• Synchronize each other’s time

• They use two or more disk shelf loops, or third-party storage, in which the following conditionsapply:

• Each node manages its own disks or array LUNs.• Each node in takeover mode manages its partner's disks or array LUNs. For third-party

storage, the partner node takes over read/write access to the array LUNs owned by the failednode until the failed node becomes available again.

Note: For systems using software-based disk ownership, disk ownership is established by DataONTAP or the administrator, rather than by which disk shelf the disk is attached to.

• They own their spare disks, spare array LUNs, or both and do not share them with the other node.• They each have mailbox disks or array LUNs on the root volume:

• Two if it is a FAS system.• One if it is a V-Series system.

The mailbox disks or LUNs are used to do the following tasks:

• Maintain consistency between the pair• Continually check whether the other node is running or whether it has performed a takeover• Store configuration information that is not specific to any particular node

• They can reside on the same Windows domain or on different domains.

Systems with variable HA configurations32xx systems can be physically configured as single-chassis active/active configurations in whichboth controller modules are in the same chassis, or as dual-chassis active/active configurations inwhich the controller modules are in separate chassis.

The following example shows a single-chassis active/active configuration:


The following example shows a dual-chassis active/active configuration and the HA interconnectcables:

Interconnect cabling for systems with variable HA configurations

Single-chassis and dual-chassis active/active configurations have different interconnect cablingrequirements.

The following table describes the interconnect cabling for 32xx systems:

If the controller modules in the active/active configurationare...

The HA interconnect cablingis...

Both in the same chassis Not required. An internalinterconnect is used.

Each in a separate chassis Required.

The HA state on systems with variable HA configurations

32xx controller modules and chassis automatically record whether they are in an active/activeconfiguration or stand-alone. This record is the HA state and must be the same on all componentswithin the stand-alone system or active/active configuration. The HA state can be manuallyconfigured if necessary.

Other systems recognize that they are in an active/active configuration by the physical presence of anNVRAM adapter in a specific slot or, for systems with an internal interconnect, the presence of a


controller module in the second bay of the chassis. Because 32xx systems can have different physicalHA configurations, they cannot rely on that method.

The HA state is recorded in the hardware PROM in the chassis and in the controller module.

The HA state can be one of the following:

ha Indicates that the controller module or chassis is in an active/active configuration.

non-ha Indicates that the controller module or chassis is in a stand-alone system.

default Indicates that the controller module or chassis has not yet been inserted in a system.After being installed using the proper procedures, the controller module or chassis isautomatically assigned the correct state depending on the state of the rest of the system.

The HA state must be consistent across all components of the system, as shown in the followingtable:

If the system or systemsare...

The HA state is recordedon these components:

The HA state on the components mustbe:

Stand-alone The chassis

Controller module A

non-ha

In a single-chassis active/active configuration

The chassis

Controller module A

Controller module B

ha

In a dual-chassis active/active configuration

Chassis A

Controller module A

Chassis B

Controller module B

ha

Related concepts

Systems with two controller modules in the same chassis on page 35

Related tasks

Verifying and setting the HA state of a 32xx controller module and chassis on page 105

If the HA state becomes inconsistent

The HA state is maintained automatically, but if for some reason it becomes inconsistent between thechassis and controller modules, an EMS message is displayed.

In that case, you can use the ha-config show and ha-config modify commands to verify andset the HA state for the controller module or modules and chassis.


Best practices for active/active configurationsTo ensure that your active/active configuration is robust and operational, you need to be familiar withconfiguration best practices.

• Make sure that the controllers and disk shelves are on different power supplies or grids, so that asingle power outage does not affect both components.

• Use VIFs (virtual interfaces) to provide redundancy and improve availability of networkcommunication.

• Follow the documented procedures in the Data ONTAP Upgrade Guide when upgrading youractive/active configuration.

• Maintain consistent configuration between the two nodes. An inconsistent configuration is oftenthe cause of failover problems.

• Test the failover capability routinely (for example, during planned maintenance) to ensure properconfiguration.

• Make sure that each node has sufficient resources to adequately support the workload of bothnodes during takeover mode.

• Use the HA Configuration Checker to help ensure that failovers are successful.• If your system supports remote management (via an RLM or Service Processor), make sure you

configure it properly, as described in the Data ONTAP System Administration Guide .• Higher numbers of traditional and FlexVol volumes on your system can affect takeover and

giveback times.When adding traditional or FlexVol volumes to an active/active configuration, consider testingthe takeover and giveback times to ensure that they fall within your requirements.

• For systems using disks, check for and remove any failed disks, as described in the Data ONTAPStorage Management Guide.

• Multipath HA is required on all active/active configurations except for some 20xx systemconfigurations which use single-path HA. Single-path HA configurations lack the redundantstandby connections.

Related tasks

Downloading and running the HA Configuration Checker utility on page 119


Comparison of active/active configuration typesThe different types of active/active configurations support different capabilities for data duplication,distance between nodes, and failover.

Active/activeconfigurationtype

Dataduplication?

Distancebetween nodes

Failoverpossible afterloss of entirenode(includingstorage)?

Notes

Standardactive/activeconfiguration

No Up to 500meters

Note: SASconfigurationsare limited to5 metersbetweennodes

No Use this configuration to providehigher availability by protectingagainst many hardware single-points-of-failure.

Mirroredactive/activeconfiguration

Yes Up to 500meters


No Use this configuration to addincreased data protection to thebenefits of a standard active/active configuration.

StretchMetroCluster

Yes Up to 500meters (270meters ifoperating at 4Gbps)


Yes Use this configuration to providedata and hardware duplication toprotect against a local disaster(for example, a power outage toone node).

MetroCluster configurations donot support SAS disk shelves.


Active/activeconfigurationtype

Dataduplication?

Distancebetween nodes

Failoverpossible afterloss of entirenode(includingstorage)?

Notes

Fabric-attachedMetroCluster

Yes Up to 100kilometers,depending onswitchconfiguration.For FASsystems, see theBrocade SwitchConfigurationGuide forFabric-attachedMetroClusters.For V-Seriessystems, up to30 km.

Yes Use this configuration to providedata and hardware duplication toprotect against a larger scaledisaster, such as the loss of anentire site.

Fabric-attached MetroClusterconfigurations do not supportSAS disk shelves.

Standard active/active configurationsStandard active/active configurations provide high availability (HA) by pairing two controllers sothat one can serve data for the other in case of controller failure or other unexpected events.

Related references

SPOF analysis for active/active configurations on page 170

How Data ONTAP works with standard active/active configurationsIn a standard active/active configuration, Data ONTAP functions so that each node monitors thefunctioning of its partner through a heartbeat signal sent between the nodes. Data from the NVRAMor NVMEM of one node is mirrored by its partner, and each node can take over the partner's disks orarray LUNs if the partner fails. Also, the nodes synchronize each other’s time.

Note: If a node reboots (but a takeover does not occur), note that the HA interconnect link comesup prior to Data ONTAP completely loading on the rebooting partner. Commands issued on thesurviving controller (that is not rebooting) that check the status of the partner or configuration mayindicate that the partner could not be reached. Wait until the partner has fully rebooted and reissuethe command.

In some cases (such as the lun config_check command) these commands are issuedautomatically when the interconnect comes up. The resulting error can generate an AutoSupport


indicating a configuration problem when in fact the underlying problem is that Data ONTAP hasnot fully booted.

Standard active/active configuration diagramA standard active/active configuration using native disk storage with multipath HA connects to thedata network, has an HA interconnect between the controllers, and has primary, standby, andredundant connections to both node's disk shelves.

Standard active/active configuration

This diagram shows a standard active/active configuration with native disk shelves andmultipath HA. Multipath HA is required on all active/active configurations except for some20xx system configurations which use single-path HA. Single-path HA configurations lack theredundant standby connections.

This diagram shows Fibre Channel disk shelves from the DS14mk2/DS14mk4 FC family. Forinformation on cabling SAS disk shelves, see the Universal SAS and ACP Cabling Guide on


the NetApp Support Site at now.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml.

Setup requirements and restrictions for standard active/activeconfigurations

You must follow certain requirements and restrictions when setting up a new standard active/activeconfiguration.

The following list specifies the requirements and restrictions to be aware of when setting up a newstandard active/active configuration:

• Architecture compatibilityBoth nodes must have the same system model and be running the same firmware version. See theData ONTAP Release Notes for the list of supported systems.

Note: In the case of systems with two controller modules in a single chassis (except the 31xxsystems), both nodes of the active/active configuration are located in the same chassis and havean internal interconnect.

• Storage capacityThe number of disks or array LUNs must not exceed the maximum configuration capacity. Ifyour system uses both native disks and third-party storage, the combined total of disks and arrayLUNs cannot exceed the maximum configuration capacity. In addition, the total storage attachedto each node must not exceed the capacity for a single node.To determine the maximum capacity for a system using disks, see the System ConfigurationGuide at http://now.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml. For asystem using array LUNs, disks, or both, see the V-Series Support Matrix.

Note: After a failover, the takeover node temporarily serves data from all the storage in theactive/active configuration. When the single-node capacity limit is less than the total active/active configuration capacity limit, the total disk space in an active/active configuration can begreater than the single-node capacity limit. It is acceptable for the takeover node to temporarilyserve more than the single-node capacity would normally allow, as long as it does not ownmore than the single-node capacity.

• Disks and disk shelf compatibility

• Fibre Channel, SATA, and SAS storage are supported in standard active/activeconfigurations, as long as the storage types are not mixed on the same loop.

• One node can have only one type of storage and the partner node can have a different type, ifneeded.

• Multipath HA is required on all active/active configurations except for some 20xx systemconfigurations which use single-path HA. Single-path HA configurations lack the redundantstandby connections.

• Cluster interconnect adapters and cables must be installed, unless the system has two controllersin the chassis and an internal interconnect.


http://support.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml


• Nodes must be attached to the same network and the Network Interface Cards (NICs) must beconfigured correctly.

• The same system software, such as Common Internet File System (CIFS), Network File System(NFS), or SyncMirror, must be licensed and enabled on both nodes.

Note: If a takeover occurs, the takeover node can provide only the functionality for the licensesinstalled on it. If the takeover node does not have a license that was being used by the partnernode to serve data, your active/active configuration loses functionality after a takeover.

Configuration variations for standard active/active configurationsActive/active configurations can be configured asymmetrically, as an active/passive pair, or withshared disk shelf stacks.

• Asymmetrical configurationsIn an asymmetrical standard active/active configuration, one node has more storage than theother. This is supported, as long as neither node exceeds the maximum capacity limit for thenode.

• Active/passive configurationsIn this configuration, the passive node has only a root volume, and the active node has all theremaining storage and services all data requests during normal operation. The passive noderesponds to data requests only if it has taken over the active node.

• Shared loops or stacksIf your standard active/active configuration is using software-based disk ownership, you canshare a loop or stack between the two nodes. This is particularly useful for active/passiveconfigurations, as described in the preceding bullet.

Multipath HA requirements and recommendationsMultipath HA is required on all active/active configurations except for some 20xx systemconfigurations which use single-path HA. Single-path HA configurations lack the redundant standbyconnections. Multipath HA was previously referred to as Multipath Storage.

What multipath HA for active/active configurations is

multipath HA provides redundancy for the path from each controller to every disk shelf in theconfiguration. It is the preferred method for cabling a storage system. An active/active configurationwithout multipath HA has only one path from each controller to every disk, but an active/activeconfiguration with multipath HA has two paths from each controller to each disk, regardless of whichnode owns the disk.

The following diagram shows the connections between the controllers and the disk shelves for anexample active/active configuration using multipath HA. The redundant primary connections and theredundant standby connections are the additional connections required for multipath HA for active/active configurations.


How the connection types are used

Outlines the connection types used for multipath HA in active/active configurations.

The following table outlines the connection types used for multipath HA for active/activeconfigurations, and how the connections are used.

Connection type How the connection is used

Primary connection For normal operation, used to serve data (load-balanced with redundant primary connection).

Redundant primary connection For normal operation, used to serve data (load-balanced with primary connection).

Standby connection For normal operation, used for heartbeatinformation only. After a takeover, assumes roleof primary connection.


Connection type How the connection is used

Redundant standby connection Not used for normal operation. After a takeover,assumes role of redundant primary connection.If the standby connection is unavailable attakeover time, assumes role of primaryconnection.

Advantages of multipath HA

Multipath connections in an active/active configuration reduce single-points-of-failure.

By providing two paths from each controller to every disk shelf, multipath HA provides thefollowing advantages:

• The loss of a disk shelf module, connection, or host bus adapter (HBA) does not require afailover. The same storage system can continue to access the data using the redundant path.

• The loss of a single disk shelf module, connection, or HBA does not prevent a successful failover.The takeover node can access its partner’s disks using the redundant path.

• You can replace modules without having to initiate a failover.

Note: While multipath HA adds value to a stretch MetroCluster environment, it is not necessary ina fabric MetroCluster configuration since multiple paths already exist.

Related concepts

Understanding redundant pathing in active/active configurations on page 150

Requirements for multipath HA

Multipath HA for active/active configurations has certain requirements.

Storage systems and disk shelves requiring multipath HA

Multipath HA is required on all active/active configurations except for some 20xx systemconfigurations which use single-path HA. Single-path HA configurations lack the redundant standbyconnections.

Disk shelf requirements

Multipath HA for active/active configurations is available on all supported disk shelves and diskmodules.

Note: Multipath HA is not supported with third-party storage connected to V-Series systems but issupported with native disks connected to V-Series systems.


Software-based disk ownership requirements

Both nodes must be using software-based disk ownership.

To convert an active/active configuration to use software-based disk ownership, you must boot bothnodes into Maintenance mode at the same time (during scheduled downtime).

Note: Plan to convert to software-based disk ownership before adding any cabling for multipathHA. After you add the cabling for multipath HA, you must manually assign all disks.

For more information about software-based disk ownership, see the chapter about disks in the DataONTAP Storage Management Guide.

Fibre Channel port requirements for Fibre Channel disk shelves

Each node must have enough onboard Fibre Channel ports or HBAs to accommodate the cablesrequired for multipath HA. You need two Fibre Channel ports for each loop.

If you are scheduling downtime to convert to software-based disk ownership, you should add theHBAs then. Otherwise, you can use the nondisruptive upgrade method to add the HBA; this methoddoes not require downtime.

Note: See System Configuration Guide at now.netapp.com for information about which slots touse for the HBAs and in what order to use them.

SAS port requirements for SAS disk shelves

Each node must have enough onboard SAS ports and/or SAS HBAs to accommodate the cablesrequired for multipath HA. You need two SAS ports on each controller for each stack.

Note: See System Configuration Guide at now.netapp.com for information about which slots touse for the HBAs and in what order to use them.

Boot environment variable requirement for V-Series systems

To use multipath HA on a V-Series system, you must configure the fc-nonarray-adapter-listenvironment variable for each new loop before you connect and configure the disk shelf formultipath HA. See the V-Series Implementation Guide for Native Disk Shelves.

Understanding mirrored active/active configurationsMirrored active/active configurations provide high availability through failover, just as standardactive/active configurations do. Additionally, mirrored active/active configurations maintain twocomplete copies of all mirrored data. These copies are called plexes and are continually and


http://now.netapp.com


synchronously updated every time Data ONTAP writes to a mirrored aggregate. The plexes can bephysically separated to protect against the loss of one set of disks or array LUNs.

Note: Mirrored active/active configurations do not provide the capability to fail over to the partnernode if one node is completely lost. For example, if power is lost to one entire node, including itsstorage, you cannot fail over to the partner node. For this capability, use a MetroCluster.

Mirrored active/active configurations use SyncMirror. For more information about SyncMirror, seethe Data ONTAP Data Protection Online Backup and Recovery Guide.

Advantages of mirrored active/active configurationsData mirroring provides additional data protection in the event of disk failures and reduces the needfor failover in the event of other component failures.

Mirroring your data protects it from the following problems which would cause data loss withoutmirroring:

• The failure or loss of two or more disks in a RAID4 aggregate• The failure or loss of three or more disks in a RAID-DP (RAID double-parity) aggregate• The failure of an array LUN; for example, because of a double disk failure on the storage array• The failure of a third-party storage array

The failure of an FC-AL adapter, SAS HBA, disk shelf loop or stack, or disk shelf module does notrequire a failover in a mirrored active/active configuration.

Similar to standard active/active configurations, if either node in a mirrored active/activeconfiguration becomes impaired or cannot access its data, the other node can automatically serve theimpaired node’s data until the problem is corrected.

Setup requirements and restrictions for mirrored active/activeconfigurations

The restrictions and requirements for mirrored active/active configurations include those for astandard active/active configuration with these additional requirements for disk pool assignments andcabling.

• You must ensure that your pools are configured correctly:

• Disks or array LUNs in the same plex must be from the same pool, with those in the oppositeplex from the opposite pool.

• If hardware-based ownership is used on your systems, the disk shelves must be connected tothe controllers so that the disks do not have to change pools when a takeover occurs.For example, on aFAS3020 system, if you connect an HBA in slot 2 to Channel A (the AInput port on the disk shelf), you should connect Channel B to an HBA that is also in pool 0on the partner node. If you connect Channel B to an HBA in, for example, slot 4, the diskswould have to change from pool 0 to pool 1 when a takeover occurs.For more information about how Data ONTAP assigns pools ownership, see the section abouthardware-based disk ownership in the Data ONTAP Storage Management Guide.


• There must be sufficient spares in each pool to account for a disk or array LUN failure.

Note: If your systems are using hardware-based disk ownership, pool membership isdetermined by the physical connections between the disk shelves and the controllers. Ifyour systems are using software-based disk ownership, pool membership is determinedexplicitly using the Data ONTAP command-line interface. For more information, see thesection on disk ownership in the Data ONTAP Storage Management Guide.

• On systems using software ownership, both plexes of a mirror should not reside on the samedisk shelf, as it would result in a single point of failure.

See the Data ONTAP Data Protection Online Backup and Recovery Guide for more informationabout requirements for setting up SyncMirror with third-party storage.

• You must enable the following licenses on both nodes:

• cluster• syncmirror_local

• If you are using third-party storage, paths to an array LUN must be redundant.

Related concepts

Setup requirements and restrictions for standard active/active configurations on page 17

Configuration variations for mirrored active/active configurationsA number of configuration variations are supported for mirrored active/active configurations.

The following list describes some configuration variations that are supported for mirrored active/active configurations:

• Asymmetrical mirroringYou can selectively mirror your storage. For example, you could mirror all the storage on onenode, but none of the storage on the other node. Takeover will function normally. However, anyunmirrored data is lost if the storage that contains it is damaged or destroyed.

Note: You must connect the unmirrored storage to both nodes, just as for mirrored storage.You cannot have storage that is connected to only one node in an active/active configuration.

Understanding stretch MetroClustersStretch MetroClusters provide data mirroring and the additional ability to initiate a failover if anentire site becomes lost or unavailable.

Like mirrored active/active configurations, stretch MetroClusters contain two complete copies of thespecified data volumes or file systems that you indicated as being mirrored volumes or file systemsin your active/active configuration. These copies are called plexes and are continually andsynchronously updated every time Data ONTAP writes data to the disks. Plexes are physicallyseparated from each other across different groupings of disks or array LUNs.


Unlike mirrored active/active configurations, MetroClusters provide the capability to force a failoverwhen an entire node (including the controllers and storage) is destroyed or unavailable.

Note: In previous versions of this document, stretch MetroClusters were called nonswitchedMetroClusters.

Note: If you are a V-Series system customer, see the V-Series MetroCluster Guide for informationabout configuring and operating a V-Series system in a MetroCluster configuration

Stretch MetroCluster configurations support SAS disk shelves with a distance limit of 5 metersbetween the nodes.

Continued data service after loss of one node with MetroClusterThe MetroCluster configuration employs SyncMirror to build a system that can continue to servedata even after complete loss of one of the nodes and the storage at that site. Data consistency isretained, even when the data is contained in more than one aggregate.

Note: You can have both mirrored and unmirrored volumes in a MetroCluster. However, theMetroCluster configuration can preserve data only if volumes are mirrored. Unmirrored volumesare lost if the storage where they reside is destroyed.

See the Data ONTAP Data Protection Online Backup and Recovery Guide for detailed informationabout using SyncMirror to mirror data.

Advantages of stretch MetroCluster configurationsMetroClusters provide the same advantages of mirroring as mirrored Active/active configurations,with the additional ability to initiate failover if an entire site becomes lost or unavailable.

For a FAS MetroCluster, the advantages of a stretch MetroCluster are:

• Your data is protected if there is a failure or loss of two or more disks in a RAID 4 aggregate orthree or more disks in a RAID-DP aggregate.

• The failure of an FC-AL adapter, loop, or ESH2 module does not require a failover.

In addition, a MetroCluster provides the cf forcetakeover -d command, giving you a singlecommand to initiate a failover if an entire site becomes lost or unavailable. If a disaster occurs at oneof the node locations and destroys your data there, your data not only survives on the other node, butcan be served by that node while you address the issue or rebuild the configuration.

Related concepts

Disaster recovery using MetroCluster on page 155


Stretch MetroCluster configuration with FAS systemsYou configure a stretch MetroCluster so that each controller can access its own storage and itspartner's storage, with local storage mirrored at the partner site.

The following figure illustrates the stretch MetroCluster configuration. The configuration includesthe following connections:

• Connections from each controller to the user network.• The MetroCluster interconnect between the two controllers.• Connections from each controller to its own storage:

• Controller A to X• Controller B to Y

• Connections from each controller to its partner's storage:

• Controller A to Y• Controller B to X

• Connections from each controller to the mirrors of its storage:

• Controller A to X-mirror• Controller B to Y-mirror


Note: This is a simplified figure that does not show disk shelf-to-disk shelf connections.

Stretch MetroCluster configuration on single enclosure active/activeconfigurations

You can configure two stretch MetroClusters between a pair of single-enclosure active/activeconfiguration systems. In this configuration, the active/active configuration between the twocontrollers in each chassis is deactivated, and two separate, side-by-side stretch MetroClusters areformed between the four controllers.

To implement the stretch MetroCluster, you must install an FC-VI adapter in each controller toprovide the cluster interconnect between the systems. When the FC-VI adapter is installed in thesystem, the internal InfiniBand interconnect is automatically disabled. This is different from otherstretch MetroClusters, which use NVRAM adapters to provide the interconnect.

The following figure shows a FAS stretch MetroCluster on single-enclosure active/activeconfiguration systems.

Note: The dual-controller FAS200 series and FAS20xx systems do not support MetroClusters.

How Data ONTAP works with stretch MetroCluster configurationsData ONTAP divides storage across physically separated pools of disks.

During configuration, Data ONTAP identifies spare disks and divides them into separate groupingscalled pools. These pools of disks are physically separated from each other, allowing for highavailability of mirrored volumes. When you add a mirrored volume or add disks to one side of amirrored volume, Data ONTAP determines how much storage you need for the second half of themirror, and dedicates that storage from a separate pool to the mirrored volume.

Data ONTAP can also be configured to read from both plexes, which in many cases improves readperformance.

Note: You can determine which side of the mirrored volume (also called a plex) is read when adata request is received using the raid.mirror_read_plex_pref option. For more information, see thena_options(1) man page.


Setup requirements and restrictions for stretch MetroClusterconfigurations

You must follow certain requirements and restrictions when setting up a new FAS StretchMetroCluster configuration.

The restrictions and requirements for stretch MetroClusters include those for a standard active/activeconfiguration and those for a mirrored active/active configuration. In addition, the followingrequirements apply:

• SAS, SATA and Fibre Channel storage is supported on stretch MetroClusters, but both plexes ofthe same aggregate must use the same type of storage.For example, you cannot mirror a Fibre Channel aggregate with SATA storage.

• Stretch MetroCluster configurations support SAS disk shelves with a distance limit of 5 metersbetween the nodes.

• MetroCluster is not supported on the FAS20xx platforms.• The following distance limitations dictate the default speed you can set:

• If the distance between the nodes is 150m and you have an 8-Gb FC-VI adapter, the defaultspeed is set to 8-Gb. If you want to increase the distance to 270m or 500m, you can set thedefault speed to 4-Gb or 2-Gb respectively.

• If the distance between nodes is between 150m and 270m and you have an 8-Gb FC-VIadapter, you can set the default speed to 4-Gb.

• If the distance between nodes is between 270m and 500m and you have an 8-Gb FC-VI or 4-Gb FC-VI adapter, you can set the default speed to 2-Gb.

• If you want to convert the stretch MetroCluster configuration to a fabric-attached MetroClusterconfiguration, you must unset the speed of the nodes before conversion by using the unsetenvcommand.

• The following licenses must be enabled on both nodes:

• cluster• syncmirror_local• cluster_remote

Note: See the MetroCluster Compatibility Matrix on the NetApp Support Site for moreinformation about hardware and firmware requirements for this configuration.

Related concepts


Setup requirements and restrictions for mirrored active/active configurations on page 22


Configuration variations for stretch MetroCluster configurationsStretch MetroClusters have asymmetrical and active/passive variations.

The following list describes some common configuration variations that are supported for stretchMetroClusters:

• Asymmetrical mirroringYou can add storage to one or both nodes that is not mirrored by the other node.

Attention: Any data contained in the unmirrored storage could be lost if that site experiences adisaster.

Note: Multiple disk failures in an unmirrored aggregate (three or more disk failures in a RAID-DP aggregate, two or more disk failures in a RAID4 aggregate) cause the node to panic,resulting in a temporary data service outage while the node reboots, a takeover occurs, ordisaster recovery is performed.

You must mirror the root volumes to enable successful takeover.


• Active/passive MetroClustersIn this configuration, the remote (passive) node does not serve data unless it has taken over forthe local (active) node. Mirroring the passive node’s root volume is optional. However, bothnodes must have all MetroCluster licenses installed so that remote takeover is possible.

Understanding fabric-attached MetroClustersLike mirrored active/active configurations, fabric-attached MetroClusters contain two complete,separate copies of the data volumes or file systems that you configured as mirrored volumes or filesystems in your active/active configuration. The fabric-attached MetroCluster nodes can bephysically distant from each other, beyond the distance limit of a stretch MetroCluster.

Fabric-attached MetroCluster configurations do not support SAS disk shelves.

Note: If you are a V-Series system customer, see the V-Series MetroCluster Guide for informationabout configuring and operating a V-Series system in a MetroCluster configuration.

Fabric-attached MetroClusters use Brocade Fibre Channel switchesA MetroCluster configuration for distances greater than 500 meters connects the two nodes usingfour Brocade Fibre Channel switches in a dual-fabric configuration for redundancy.

Each site has two Fibre Channel switches, each of which is connected through an inter-switch link toa partner switch at the other site. The inter-switch links are fiber optic connections that provide agreater distance between nodes than other active/active configurations. For FAS systems, see the


Brocade Switch Configuration Guide for Fabric-attached MetroClusters and for V-Series system, seethe V-Series Support Matrix.

Each local switch combines with a partner switch to form a fabric. By using four switches instead oftwo, redundancy is provided to avoid single-points-of-failure in the switches and their connections.

Like a stretch MetroCluster configuration, a fabric-attached MetroCluster employs SyncMirror tobuild a system that can continue to serve data even after complete loss of one of the nodes and thestorage at that site. Data consistency is retained, even when the data is contained in more than oneaggregate.

Related information

Fabric-attached MetroCluster Brocade Switch Configuration Guide - https://now.netapp.com/NOW/knowledge/docs/docs.cgi

Advantages of fabric-attached MetroCluster configurationsFabric-attached MetroClusters provide the same advantages of stretch MetroCluster configurations,while also enabling the physical nodes to be physically distant from each other.

The advantages of a fabric-attached MetroCluster over a stretch MetroCluster include the following:

• The two halves of the configuration can be more than 500 meters apart, which provides increaseddisaster protection.

• Disk shelves and nodes are not connected directly to each other, but are connected to a fabric withmultiple data routes, ensuring no single point of failure.

Fabric-attached MetroCluster configuration for FAS systemsA fabric-attached MetroCluster includes two Brocade Fibre Channel switch fabrics that provide longdistance connectivity between the nodes. Through the Brocade switches, each controller can accessits own storage and its partner's storage, with local storage mirrored at the partner site.

The following figure illustrates the fabric-attached MetroCluster configuration.


https://now.netapp.com/NOW/knowledge/docs/docs.cgi



Fabric-attached MetroCluster configuration on single-enclosure active/active configurations

You can configure a fabric-attached MetroCluster between a pair of single-enclosure active/activeconfiguration systems. In this configuration, the active/active configuration between the twocontrollers in each chassis is deactivated, and two separate, side-by-side MetroClusters are formedbetween the four controllers.

When the system detects the presence of an FC-VI adapter, which connects the controller to theBrocade switch fabric, the internal InfiniBand connection is automatically deactivated.

The following figure shows a FAS fabric-attached MetroCluster on single-enclosure systems.



How Data ONTAP works with fabric-attached MetroCluster configurationsData ONTAP functions the same way on a fabric-attached MetroCluster as on a stretch MetroCluster.

Related concepts

How Data ONTAP works with stretch MetroCluster configurations on page 26

Setup requirements and restrictions for fabric-attached MetroClustersYou must follow certain requirements and restrictions when setting up a new fabric-attachedMetroCluster configuration.

The setup requirements for a fabric-attached MetroCluster include those for standard and mirroredActive/active configurations, with the following exceptions:


Note: See the MetroCluster Compatibility Matrix on the NetApp Support Site for moreinformation about hardware and firmware requirements for this configuration.

Node requirements

• The nodes must be one of the following system models configured for mirrored volume use; eachnode in the pair must be the same model.

• FAS900 series systems• FAS30xx systems• 31xx systems• The 31xx systems can have two controllers in the same chassis.

When in a MetroCluster configuration, only a single controller in each system is supported(rather than two).

• 60xx• 32xx systems

• Each node requires a FC-VI (Fibre Channel-Virtual Interface) adapter; the slot position isdependent on the controller model.The 4-Gbps FC-VI adapter is supported on the following systems using software-based diskownership:

• FAS60xx• FAS3040• FAS3070• 31xx

Note: For information about supported cards and slot placement, see the System ConfigurationGuide on the NetApp Support Site.

The FC-VI adapter is also called a VI-MC or VI-MetroCluster adapter.

• The 8-Gbps FC-VI (Fibre Channel-Virtual Interface) adapter is supported only on the 32xxsystems.

Disk and disk shelf requirements

• Only DS14mk2 and DS14mk4 FC disk shelves are supported.• Only Fibre Channel disks are supported; you cannot use SATA drives or AT-FCX modules for

fabric-attached MetroCluster configurations.• Only homogeneous stacks of SAS disk shelves are supported. For example, a stack of disk

shelves can only contain either DS4243 or DS2246 disk shelves.• You can connect a maximum of two DS14mk2 disk shelves to each loop.


Capacity limits

The maximum capacity for a system configured in a fabric-attached MetroCluster is the smallest ofthe following limits:

• The maximum storage capacity for the node.

Note: For the maximum storage capacity, see the System Configuration Guide on the NetAppSupport Site.

• 840 Fibre Channel disks (60 disk shelves).

Fibre Channel switch requirements

Note: For the most up-to-date switch information, including supported switches and firmwaredownloads, see the Fabric-Attached MetroCluster Switch Description page on the NetApp SupportSite. (To access this page, navigate to Download Software > Fibre Channel Switch > Brocade.)

• Each site of the MetroCluster requires two switches.• Switches must be a supported Brocade model supplied by NetApp. Customer supplied switches

are not supported.• The two switches at a site must be the same model and must be licensed for the same number of

ports.• All the four switches for a particular Fabric-attached MetroCluster must support the same

maximum speed.• Switches must be running the correct firmware version.

License requirements

• cluster• syncmirror_local• cluster_remote

Related concepts


Setup requirements and restrictions for mirrored active/active configurations on page 22

Configuration limitations for fabric-attached MetroClustersYou must be aware of certain limitations when setting up a new fabric-attached MetroClusterconfiguration.

The fabric-attached MetroCluster configuration has the following limitations:

• SATA and AT-FCX storage is not supported.• You cannot use the MetroCluster switches to connect Fibre Channel tape devices, or for Fibre

Channel Protocol (FCP) traffic of any kind.You can connect only system controllers and disk shelves to the MetroCluster switches.


• You can connect a tape storage area network (SAN) to either of the nodes, but the tape SAN mustnot use the MetroCluster switches.

Configuration variations for fabric-attached MetroClustersFabric-attached MetroClusters support asymmetrical and active/passive configurations.

The following list describes some common configuration variations that are supported for fabric-attached MetroClusters:

• Asymmetrical mirroringYou can add storage to one or both nodes that is not mirrored by the other node. However, anydata contained in the unmirrored storage could be lost if that site experiences a disaster.

Attention: Multiple disk failures in an unmirrored aggregate (three or more disk failures in aRAID-DP aggregate, two or more disk failures in a RAID4 aggregate) will cause the node topanic, resulting in a temporary data service outage while the node reboots or disaster recoveryis performed.

You must mirror the root volumes to enable successful takeover.


• Active/passive MetroClustersIn this configuration, the remote (passive) node does not serve data unless it has taken over forthe local (active) node. Mirroring the passive node’s root volume is optional. However, bothnodes must have all MetroCluster licenses installed so that remote takeover is possible.


Active/active configuration installation

To install and cable a new standard or mirrored active/active configuration you must have the correcttools and equipment and you must connect the controllers to the disk shelves (for FAS systems or V-Series systems using native disk shelves). You must also cable the cluster interconnect between thenodes. Active/active configurations can be installed in either NetApp system cabinets or inequipment racks.

The specific procedure you use depends on the following aspects of your configuration:

• Whether you have a standard or mirrored active/active configuration.• Whether you are using Fibre Channel or SAS disk shelves.

Note: If your configuration includes SAS disk shelves, see the Universal SAS and ACPCabling Guide on the NetApp Support Site at support.netapp.com/ for information on diskshelf cabling. For cabling the HA interconnect between the nodes, use the procedures in thisguide.

Multipath HA is required on all active/active configurations except for some 20xx systemconfigurations which use single-path HA. Single-path HA configurations lack the redundant standbyconnections.

Systems with two controller modules in the same chassisIn an active/active configuration, some storage systems (such as the 20xx systems) support twocontroller modules in the same chassis.

This simplifies system cabling, because an internal InfiniBand connector between the two controllermodules replaces the interconnect adapters and cabling used in other configurations. The followingillustration shows a 32xx system with two controller modules in the chassis forming a single chassisactive/active configuration.

This configuration is different from other examples in this section, which show systems that must becabled together with an interconnect to enable the active/active configuration.

Related concepts

Systems with variable HA configurations on page 10

Active/active configuration installation | 35

http://support.netapp.com/

System cabinet or equipment rack installationYou need to install your active/active configuration in one or more NetApp system cabinets or instandard telco equipment racks. Each of these options has different requirements.

Active/active configurations in an equipment rackDepending on the amount of storage you ordered, you need to install the equipment in one or moretelco-style equipment racks.

The equipment racks can hold one or two nodes on the bottom and eight or more disk shelves. Forinformation about how to install the disk shelves and nodes into the equipment racks, see theappropriate documentation that came with your equipment.

Active/active configurations in a system cabinetIf you ordered an active/active configuration in a system cabinet, it comes in one or more NetAppsystem cabinets, depending on the amount of storage.

The number of system cabinets you receive depends on how much storage you ordered. All internaladapters, such as networking adapters, Fibre Channel adapters, and other adapters, arrive preinstalledin the nodes.

If the active/active configuration you ordered has six or fewer disk shelves, it arrives in a singlesystem cabinet. This system cabinet has both the Channel A and Channel B disk shelves cabled, andalso has the cluster adapters cabled.

If the active/active configuration you ordered has more than six disk shelves, the active/activeconfiguration arrives in two or more system cabinets. You must complete the cabling by cabling thelocal node to the partner node’s disk shelves and the partner node to the local node’s disk shelves.You must also cable the nodes together by cabling the NVRAM cluster interconnects. If the active/active configuration uses switches, you must install the switches, as described in the accompanyingswitch documentation. The system cabinets might also need to be connected to each other. See yourSystem Cabinet Guide for information about connecting your system cabinets together.

Required documentation, tools, and equipmentInstallation of an active/active configuration requires the correct documentation, tools, andequipment.

Required documentationInstallation of an active/active configuration requires the correct documentation.

NetApp hardware and service documentation is not contained within a single guide. Instead, thefield-replaceable units are documented in separate flyers at the NetApp Support Site.


The following table lists and briefly describes the documentation you might need to refer to whenpreparing a new active/active configuration, or converting two stand-alone systems into an active/active configuration.

Manual name Description

The appropriate system cabinet guide This guide describes how to install NetAppequipment into a system cabinet.

Site Requirements Guide This guide describes the physical requirementsyour site must meet to install NetAppequipment.

The appropriate disk shelf guide These guides describe how to cable a disk shelfto a storage system.

The appropriate hardware documentation foryour storage system model

These guides describe how to install the storagesystem, connect it to a network, and bring it upfor the first time.

Diagnostics Guide This guide describes the diagnostics tests thatyou can run on the storage system.

Data ONTAP Upgrade Guide This guide describes how to upgrade storagesystem and disk firmware, and how to upgradestorage system software.

Data ONTAP Data Protection Online Backupand Recovery Guide

This guide describes, among other topics,SyncMirror technology, which is used formirrored active/active configurations.

Data ONTAP System Administration Guide This guide describes general storage systemadministration.

Data ONTAP Software Setup Guide This guide describes how to configure thesoftware of a new storage system for the firsttime.

Note: If you are installing a V-Series active/active configuration, refer also to the V-SeriesInstallation Requirements and Reference Guide for information about cabling V-Series systems tostorage arrays and to the V-Series Implementation Guides for information about configuringstorage arrays to work with V-Series systems.

Related information

Data ONTAP Information Library - http://now.netapp.com/NOW/knowledge/docs/ontap/ontap_index.shtml


http://now.netapp.com/NOW/knowledge/docs/ontap/ontap_index.shtml


Required toolsInstallation of an active/active configuration requires the correct tools.

The following list specifies the tools you need to install the active/active configuration:

• #1 and #2 Phillips screwdrivers• Hand level• Marker

Required equipmentWhen you receive your active/active configuration, you should receive the equipment listed in thefollowing table. See the System Configuration Guide at the NetApp Support Site to confirm yourstorage system type, storage capacity, and so on.

Required equipment Standard or mirrored active/activeconfiguration

Storage system Two of the same type of storage systems.

Storage See the System Configuration Guide at theNetApp Support Site.

Cluster interconnect adapter (for controllermodules that do not share a chassis)

Note: When 32xx systems are in a dual-chassis active/active configuration, they usethe c0a and c0b 10-GbE ports for the HAinterconnect. They do not require an HAinterconnect adapter.

InfiniBand (IB) cluster adapter

(The NVRAM adapter functions as the clusterinterconnect adapter on FAS900 series and laterstorage systems, except the 32xx systems.)

For DS14mk2/DS14mk4 FC family diskshelves: FC-AL or FC HBA (FC HBA for Disk)adapters

For SAS disk shelves: SAS HBAs, if applicable

Minimum of two FC-AL adapters or two SASHBAs

Fibre Channel switches N/A

SFP (Small Form Pluggable) modules N/A

NVRAM cluster adapter media converter Only if using fiber cabling.


Required equipment Standard or mirrored active/activeconfiguration

Cables (provided with shipment unlessotherwise noted)

• One optical controller-to-disk shelf cable perloop

• Multiple disk shelf-to-disk shelf cables• Two 4xIB copper cables, or two 4xIB optical

cables

Note: You must purchase longer opticalcables separately for cabling distancesgreater than 30 meters.

• Two optical cables with media converters forsystems using the IB cluster adapter

• The 32xx systems, when in a dual-chassisactive/active configuration, require 10 GbEcables (Twinax or SR) for the HAinterconnect.

Preparing your equipmentYou must install your nodes in your system cabinets or equipment racks, depending on yourinstallation type.

Installing the nodes in equipment racksBefore you cable your nodes together, you install the nodes and disk shelves in the equipment rack,label the disk shelves, and connect the nodes to the network.

Steps

1. Install the nodes in the equipment rack, as described in the guide for your disk shelf, hardwaredocumentation, or Quick Start guide that came with your equipment.

2. Install the disk shelves in the equipment rack, as described in the appropriate disk shelf guide.

3. Label the interfaces, where appropriate.

4. Connect the nodes to the network, as described in the setup instructions for your system.

Result

The nodes are now in place and connected to the network and power is available.


After you finish

Proceed to cable the active/active configuration.

Installing the nodes in a system cabinetBefore you cable your nodes together, you must install the system cabinet and connect the nodes tothe network. If you have two cabinets, the cabinets must be connected together.

Steps

1. Install the system cabinets, as described in the System Cabinet Guide. If you have multiplesystem cabinets, remove the front and rear doors and any side panels that need to be removed,and connect the system cabinets together.

2. Connect the nodes to the network.

3. Connect the system cabinets to an appropriate power source and apply power to the cabinets.

Result

The nodes are now in place and connected to the network and power is available.

After you finish

Proceed to cable the active/active configuration.

Cabling a standard active/active configurationTo cable a standard active/active configuration, you identify the ports you need to use on each node,then you cable the ports, and then you cable the cluster interconnect.

About this task

This procedure explains how to cable a configuration using DS14mk2 or DS14mk4 FC disk shelves.

For cabling SAS disk shelves in an active/active configuration, see the Universal SAS and ACPCabling Guide.

The sections for cabling the cluster interconnect apply to all systems regardless of disk shelf type.

Steps

1. Determining which Fibre Channel ports to use for Fibre Channel disk shelfconnections on page 41

2. Cabling Node A to DS14mk2 or DS14mk4 FC disk shelves on page 42

3. Cabling Node B to DS14mk2 or DS14mk4 FC disk shelves on page 44

4. Cabling the cluster interconnect (all systems except 32xx) on page 46


5. Cabling the cluster interconnect (32xx systems in separate chassis) on page 47

Determining which Fibre Channel ports to use for Fibre Channel disk shelfconnections

Before cabling your active/active configuration, you need to identify which Fibre Channel ports touse to connect your disk shelves to each storage system, and in what order to connect them.

Keep the following guidelines in mind when identifying ports to use:

• Every disk shelf loop in the active/active configuration requires two ports on the node, one for theprimary connection and one for the redundant multipath HA connection.A standard active/active configuration with one loop for each node uses four ports on each node.

• Onboard Fibre Channel ports should be used before using ports on expansion adapters.• Always use the expansion slots in the order shown in the System Configuration Guide at the

NetApp Support Site for your platform for an active/active configuration.• If using Fibre Channel HBAs, insert the adapters in the same slots on both systems.

See the System Configuration Guide at http://now.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml to obtain all slot assignment information for the various adapters you use tocable your active/active configuration.

When complete, you should have a numbered list of Fibre Channel ports for both nodes, starting withPort 1.

Cabling guidelines for a quad-port Fibre Channel HBA

If using ports on the quad-port, 4-Gb Fibre Channel HBAs, use the procedures in the followingsections, with the following additional guidelines:

• Disk shelf loops using ESH4 modules must be cabled to the quad-port HBA first.• Disk shelf loops using AT-FCX or ESH2 modules must be cabled to dual-port HBA ports or

onboard ports before using ports on the quad-port HBA.• Port A of the HBA must be cabled to the In port of Channel A of the first disk shelf in the loop.

Port A of the partner node's HBA must be cabled to the In port of Channel B of the first disk shelfin the loop. This ensures that disk names are the same for both nodes.

• Additional disk shelf loops must be cabled sequentially with the HBA’s ports. Use port A for thefirst loop, then B, and so on.

• If available, ports C or D must be used for the redundant multipath HA connection after cablingall remaining disk shelf loops.

• All other cabling rules described in the documentation for the HBA and the System ConfigurationGuide must be observed.


Cabling Node A to DS14mk2 or DS14mk4 FC disk shelvesTo cable Node A, you must use the Fibre Channel ports you previously identified and cable the diskshelf loops owned by the node to these ports.

About this task

• This procedure uses multipath HA, required on all systems.• This procedure does not apply to SAS disk shelves.


• For additional cabling diagrams, you can refer to your system's Installation and Setup Instructionson the NetApp Support Site.

Steps

1. Review the cabling diagram before proceeding to the cabling steps.

• The circled numbers in the diagram correspond to the step numbers in the procedure.• The location of the Input and Output ports on the disk shelves vary depending on the disk

shelf models.Make sure that you refer to the labeling on the disk shelf rather than to the location of the portshown in the diagram.

• The location of the Fibre Channel ports on the controllers is not representative of anyparticular storage system model; determine the locations of the ports you are using in yourconfiguration by inspection or by using the Installation and Setup Instructions for your model.

• The port numbers refer to the list of Fibre Channel ports you created.• The diagram only shows one loop per node and one disk shelf per loop.

Your installation might have more loops, more disk shelves, or different numbers of diskshelves between nodes.


2. Cable Fibre Channel port A1 of Node A to the Channel A Input port of the first disk shelf ofNode A loop 1.

3. Cable the Node A disk shelf Channel A Output port to the Channel A Input port of the next diskshelf in loop 1.

4. Repeat step 3 for any remaining disk shelves in loop 1.

5. Cable the Channel A Output port of the last disk shelf in the loop to Fibre Channel port B4 ofNode B.

This provides the redundant multipath HA connection for Channel A.

6. Cable Fibre Channel port A2 of Node A to the Channel B Input port of the first disk shelf ofNode B loop 1.

7. Cable the Node B disk shelf Channel B Output port to the Channel B Input port of the next diskshelf in loop 1.


9. Cable the Channel B Output port of the last disk shelf in the loop to Fibre Channel port B3 ofNode B.

This provides the redundant multipath HA connection for Channel B.

10. Repeat steps 2 to 9 for each pair of loops in the active/active configuration, using ports 3 and 4for the next loop, ports 5 and 6 for the next one, and so on.

Result

Node A is completely cabled.


After you finish

Proceed to cabling Node B.

Cabling Node B to DS14mk2 or DS14mk4 FC disk shelvesTo cable Node B, you must use the Fibre Channel ports you previously identified and cable the diskshelf loops owned by the node to these ports.

About this task



• For additional cabling diagrams, you can refer to your system's Installation and Setup Instructionson the NetApp Support Site.

Steps








2. Cable Port B1 of Node B to the Channel B Input port of the first disk shelf of Node A loop 1.

Both channels of this disk shelf are connected to the same port on each node. This is not required,but it makes your active/active configuration easier to administer because the disks have the sameID on each node. This is true for Step 5 also.

3. Cable the disk shelf Channel B Output port to the Channel B Input port of the next disk shelf inloop 1.


5. Cable the Channel B Output port of the last disk shelf in the loop to Fibre Channel port A4 ofNode A.

This provides the redundant multipath HA connection for Channel B.

6. Cable Fibre Channel port B2 of Node B to the Channel A Input port of the first disk shelf of NodeB loop 1.

7. Cable the disk shelf Channel A Output port to the Channel A Input port of the next disk shelf inloop 1.


9. Cable the Channel A Output port of the last disk shelf in the loop to Fibre Channel port A3 ofNode A.

This provides the redundant multipath HA connection for Channel A.


10. Repeat steps 2 to 9 for each pair of loops in the active/active configuration, using ports 3 and 4for the next loop, ports 5 and 6 for the next one, and so on.

Result

Node B is completely cabled.

After you finish

Proceed to cable the cluster interconnect.

Cabling the cluster interconnect (all systems except 32xx)To cable the cluster interconnect between the active/active configuration nodes, you must make surethat your interconnect adapter is in the correct slot and connect the adapters on each node with theoptical cable.

About this task

This procedure applies to all dual-chassis active/active configurations (active/active configurations inwhich the two controller modules reside in separate chassis) except 32xx systems, regardless of diskshelf type.

Steps

1. See the System Configuration Guide on the NetApp Support Site to ensure that your interconnectadapter is in the correct slot for your system in an active/active configuration.

For systems that use an NVRAM adapter, the NVRAM adapter functions as the clusterinterconnect adapter.

2. Plug one end of the optical cable into one of the local node's cluster adapter ports, then plug theother end into the partner node's corresponding adapter port.

You must not cross-cable the cluster interconnect adapter. Cable the local node ports only to theidentical ports on the partner node.

If the system detects a cross-cabled cluster interconnect, the following message appears:

Cluster interconnect port <port> of this appliance seems to be connectedto port <port> on the partner appliance.

3. Repeat Step 2 for the two remaining ports on the cluster adapters.

Result

The nodes are connected to each other.

After you finish

Proceed to configure the system.


Cabling the cluster interconnect (32xx systems in separate chassis)To enable the cluster interconnect between 32xx controller modules that reside in separate chassis,you must cable the onboard 10-GbE ports on one controller module to the onboard GbE ports on thepartner.

About this task

This procedure applies to 32xx systems regardless of the type of attached disk shelves.

Steps

1. Plug one end of the 10-GbE cable to the c0a port on one controller module.

2. Plug the other end of the 10-GbE cable to the c0a port on the partner controller module.

3. Repeat the preceding steps to connect the c0b ports.

Do not cross-cable the HA interconnect adapter; cable the local node ports only to the identicalports on the partner node.

Result


After you finish


Cabling a mirrored active/active configurationTo cable a mirrored active/active configuration, you identify the ports you need to use on each node,and then you cable the ports, and then you cable the cluster interconnect.

About this task

This procedure explains how to cable a configuration using DS14mk2 or DS14mk4 FC disk shelves.


The sections for cabling the cluster interconnect apply to all systems regardless of disk shelf type.

Steps

1. Determining which Fibre Channel ports to use for Fibre Channel disk shelfconnections on page 41

2. Creating your port list for mirrored active/active configurations on page 49


3. Cabling the channel A DS14mk2 or DS14mk4 FC disk shelf loops on page 50

4. Cabling the channel B DS14mk2 or DS14mk4 FC disk shelf loops on page 52

5. Cabling the redundant multipath HA connection for each loop on page 54

6. Cabling the cluster interconnect (all systems except 32xx) on page 56

7. Cabling the cluster interconnect (32xx systems in separate chassis) on page 57

Determining which Fibre Channel ports to use for Fibre Channel disk shelfconnections

Before cabling your active/active configuration, you need to identify which Fibre Channel ports touse to connect your disk shelves to each storage system, and in what order to connect them.

Keep the following guidelines in mind when identifying ports to use:

• Every disk shelf loop in the active/active configuration requires two ports on the node, one for theprimary connection and one for the redundant multipath HA connection.A standard active/active configuration with one loop for each node uses four ports on each node.

• Onboard Fibre Channel ports should be used before using ports on expansion adapters.• Always use the expansion slots in the order shown in the System Configuration Guide at the

NetApp Support Site for your platform for an active/active configuration.• If using Fibre Channel HBAs, insert the adapters in the same slots on both systems.

See the System Configuration Guide at http://now.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml to obtain all slot assignment information for the various adapters you use tocable your active/active configuration.

When complete, you should have a numbered list of Fibre Channel ports for both nodes, starting withPort 1.

Cabling guidelines for a quad-port Fibre Channel HBA

If using ports on the quad-port, 4-Gb Fibre Channel HBAs, use the procedures in the followingsections, with the following additional guidelines:

• Disk shelf loops using ESH4 modules must be cabled to the quad-port HBA first.• Disk shelf loops using AT-FCX or ESH2 modules must be cabled to dual-port HBA ports or

onboard ports before using ports on the quad-port HBA.• Port A of the HBA must be cabled to the In port of Channel A of the first disk shelf in the loop.

Port A of the partner node's HBA must be cabled to the In port of Channel B of the first disk shelfin the loop. This ensures that disk names are the same for both nodes.

• Additional disk shelf loops must be cabled sequentially with the HBA’s ports. Use port A for thefirst loop, then B, and so on.

• If available, ports C or D must be used for the redundant multipath HA connection after cablingall remaining disk shelf loops.

• All other cabling rules described in the documentation for the HBA and the System ConfigurationGuide must be observed.


Creating your port list for mirrored active/active configurationsAfter you determine the Fibre Channel ports to use, you create a table identifying which ports belongto which port pool.

About this task

Mirrored active/active configurations, regardless of disk shelf type, use SyncMirror to separate eachaggregate into two plexes that mirror each other. One plex uses disks in pool 0 and the other plexuses disks in pool 1. To ensure proper disk pool access, your cabling depends on whether you havehardware-based or software-based disk ownership.

If your system uses hardware-based disk ownership, you must cable your mirrored active/activeconfiguration according to the pool rules for your platform. For more information about pool rules,see the section on hardware-based disk ownership in the Data ONTAP Storage Management Guide.

If your system uses software-based disk ownership, follow the guidelines for software-based diskownership in the Data ONTAP Storage Management Guide.

For more information about SyncMirror, see the Data ONTAP Data Protection Online Backup andRecovery Guide.

Step

1. Create a table specifying the port usage; the cabling diagrams in this document use the notation“P1-3” (the third port for pool 1).

Example

For a30xx active/active configuration that has two mirrored loops using hardware-based diskownership, the port list would look like the following example:

Pool 0 Pool 1

P0-1: onboard port 0a P1-1: onboard port 0c

P0-2: onboard port 0b P1-2: onboard port 0d

P0-3: slot 2 port A P1-3: slot 4 port A

P0-4: slot 2 port B P1-4: slot 4 port B

After you finish

Proceed to cable the Channel A loops.


Cabling the channel A DS14mk2 or DS14mk4 FC disk shelf loopsTo begin cabling of the disk shelves, you cable the appropriate pool ports on the node to the ChannelA modules of the disk shelf stack for the pool.

About this task



Steps

1. Complete your port list.








3. Cable channel A for Node A.

a. Cable the first port for pool 0 (P0-1) of Node A to the first Node A disk shelf Channel A Inputport of disk shelf pool 0.

b. Cable the first port for pool 1 (P1-1) of Node A to the first Node A disk shelf Channel A Inputport of disk shelf pool 1.

c. Cable the disk shelf Channel A Output port to the next disk shelf Channel A Input port in theloop for both disk pools.


Note: The illustration shows only one disk shelf per disk pool. The number of disk shelvesper pool might be different for your configuration.

d. Repeat substep 2c, connecting Channel A output to input, for any remaining disk shelves inthis loop for each disk pool.

e. Repeat Substep a through Substep e for any additional loops for Channel A, Node A, usingthe odd numbered port numbers (P0-3 and P1-3, P0-5 and P1-5, and so on).

4. Cable Channel A for Node B

a. Cable the second port for pool 0 (P0-2) of Node B to the first Node B disk shelf Channel AInput port of disk shelf pool 0.

b. Cable the second port for pool 1 (P1-2) of Node B to the first Node B disk shelf Channel AInput port of disk shelf pool 1.

c. Cable the disk shelf Channel A Output port to the next disk shelf Channel A Input port in theloop for both disk pools.

d. Repeat substep 3.c, connecting Channel A output to input, for any remaining disk shelves ineach disk pool.

e. Repeat substep 3.a through substep 3.e for any additional loops on Channel A, Node B, usingthe even numbered port numbers (P0-4 and P1-4, P0-6 and P1-6, and so on).

After you finish

Proceed to cable the Channel B loops.

Cabling the channel B DS14mk2 or DS14mk4 FC disk shelf loopsTo provide the mirrored storage, you cable the mirrored pool ports on the node to the Channel Bmodules of the appropriate disk shelf stack.

Steps








2. Cable Channel B for Node A.

a. Cable the second port for pool 0 (P0-2) of Node A to the first Node B disk shelf Channel BInput port of disk shelf pool 0.


Note: Both channels of this disk shelf are connected to the same port on each node. This isnot required, but it makes your active/active configuration easier to administer because thedisks have the same ID on each node.

b. Cable the second port for pool 1 (P1-2) of Node A to the first Node B disk shelf Channel BInput port of disk shelf pool 1.

c. Cable the disk shelf Channel B Output port to the next disk shelf Channel B Input port in theloop for both disk pools.

Note: The illustration shows only one disk shelf per disk pool. The number of disk shelvesper pool might be different for your configuration.

d. Repeat Substep c, connecting Channel B output to input, for any remaining disk shelves ineach disk pool.

e. Repeat Substep a through Substep d for any additional loops on Channel B, Node A, using theeven numbered port numbers (P0-4 and P1-4, P0-6 and P1-6, and so on).

3. Cable Channel B for Node B.

a. Cable the first port for pool 0 (P0-1) of Node B to the first Node A disk shelf Channel B Inputport of disk shelf pool 0.

b. Cable the first port for pool 1 (P1-1) of Node B to the first Node A disk shelf Channel B Inputport of disk shelf pool 1.

c. Cable the disk shelf Channel B Output port to the next disk shelf Channel B Input port in theloop for both disk pools.

d. Repeat Substep c, connecting Channel B output to input, for any remaining disk shelves ineach disk pool.

e. Repeat Substep a through Substep d for any additional loops for Channel B, Node B, using theodd numbered port numbers (P0-3 and P1-3, P0-5 and P1-5, and so on).

After you finish

Proceed to cable the cluster interconnect.

Cabling the redundant multipath HA connection for each loopTo complete the multipath HA cabling for the disk shelves, you must add the final connection foreach channel on the final disk shelf in each loop.

Steps



shelf models.


Make sure that you refer to the labeling on the disk shelf rather than to the location of the portshown in the diagram.





2. Connect the Channel A output port on the last disk shelf for each loop belonging to Node A to anavailable port on Node B in the same pool.

3. Connect the Channel B output port on the last disk shelf for each loop belonging to Node A to anavailable port on Node B in the same pool.

4. Connect the Channel A output port on the last disk shelf for each loop belonging to Node B to anavailable port on Node B in the same pool.

5. Connect the Channel B output port on the last disk shelf for each loop belonging to Node B to anavailable port on Node B in the same pool.

Cabling the cluster interconnect (all systems except 32xx)To cable the cluster interconnect between the active/active configuration nodes, you must make surethat your interconnect adapter is in the correct slot and connect the adapters on each node with theoptical cable.

About this task

This procedure applies to all dual-chassis active/active configurations (active/active configurations inwhich the two controller modules reside in separate chassis) except 32xx systems, regardless of diskshelf type.

Steps

1. See the System Configuration Guide on the NetApp Support Site to ensure that your interconnectadapter is in the correct slot for your system in an active/active configuration.

For systems that use an NVRAM adapter, the NVRAM adapter functions as the clusterinterconnect adapter.

2. Plug one end of the optical cable into one of the local node's cluster adapter ports, then plug theother end into the partner node's corresponding adapter port.

You must not cross-cable the cluster interconnect adapter. Cable the local node ports only to theidentical ports on the partner node.

If the system detects a cross-cabled cluster interconnect, the following message appears:

Cluster interconnect port <port> of this appliance seems to be connectedto port <port> on the partner appliance.

3. Repeat Step 2 for the two remaining ports on the cluster adapters.

Result


After you finish



Cabling the cluster interconnect (32xx systems in separate chassis)To enable the cluster interconnect between 32xx controller modules that reside in separate chassis,you must cable the onboard 10-GbE ports on one controller module to the onboard GbE ports on thepartner.

About this task

This procedure applies to 32xx systems regardless of the type of attached disk shelves.

Steps

1. Plug one end of the 10-GbE cable to the c0a port on one controller module.

2. Plug the other end of the 10-GbE cable to the c0a port on the partner controller module.

3. Repeat the preceding steps to connect the c0b ports.

Do not cross-cable the HA interconnect adapter; cable the local node ports only to the identicalports on the partner node.

Result


After you finish


Required connections for using uninterruptible powersupplies with standard or mirrored active/activeconfigurations

You can use a UPS (uninterruptible power supply) with your active/active configuration. The UPSenables the system to fail over gracefully if power fails for one of the nodes, or to shut downgracefully if power fails for both nodes. You must ensure that the correct equipment is connected tothe UPS.

To gain the full benefit of the UPS, you must ensure that all the required equipment is connected tothe UPS. The equipment that needs to be connected depends on whether your configuration is astandard or a mirrored active/active configuration.

For a standard active/active configuration, you must connect the controller, disks, and any FibreChannel switches in use.

For a mirrored active/active configuration, you must connect the controller and any Fibre Channelswitches to the UPS, as for a standard active/active configuration. However, if the two sets of diskshelves have separate power sources, you do not have to connect the disks to the UPS. If power is


interrupted to the local controller and disks, the controller can access the remote disks until it shutsdown gracefully or the power supply is restored. In this case, if power is interrupted to both sets ofdisks at the same time, the active/active configuration cannot shut down gracefully.


MetroCluster installation

You can install a stretch or fabric-attached MetroCluster to provide complete data mirroring andtakeover capabilities if a site is lost in a disaster. Fabric-attached MetroClusters provide active/activeconfiguration with physically separated nodes at a greater distance than that provided by stretchMetroCluster.


Related concepts


Setup requirements and restrictions for stretch MetroCluster configurations on page 27

Setup requirements and restrictions for fabric-attached MetroClusters on page 31

Required documentation, tools, and equipmentDescribes the NetApp documentation and the tools required to install a MetroCluster configuration.

Required documentationDescribes the flyers and guides required to install a new MetroCluster, or convert two stand-alonesystems into a MetroCluster.

NetApp hardware and service documentation is not contained within a single guide. Instead, thefield-replaceable units are documented in separate flyers at the NetApp Support Site.

The following table lists and briefly describes the documentation you might need to refer to whenpreparing a new MetroCluster configuration, or converting two stand-alone systems into aMetroCluster configuration.


The appropriate system cabinet guide This guide describes how to install NetAppequipment into a system cabinet.

Site Requirements Guide This guide describes the physical requirementsyour site must meet to install NetAppequipment.

The appropriate disk shelf guide These guides describe how to cable a disk shelfto a storage system.

MetroCluster installation | 59


The appropriate hardware documentation foryour storage system model

These guides describe how to install the storagesystem, connect it to a network, and bring it upfor the first time.

Diagnostics Guide This guide describes the diagnostics tests thatyou can run on the storage system.

Upgrade Guide This guide describes how to upgrade storagesystem and disk firmware, and how to upgradestorage system software.

Data Protection Online Backup and RecoveryGuide

This guide describes, among other topics,SyncMirror technology, which is used formirrored Active/active configurations.

Data ONTAP System Administration Guide This guide describes general storage systemadministration.

Software Setup Guide This guide describes how to configure thesoftware of a new storage system for the firsttime.

Brocade Switch Configuration Guide for Fabric-attached MetroClusters

This document describes how to configureBrocade switches for a fabric-attachedMetroCluster.

You can find this document on the Fabric-attached MetroCluster Switch Description pageon the NetApp Support Site.

The appropriate Brocade manuals These guides describe how to configure andmaintain Brocade switches.

These guides are available from the BrocadeSwitch Description Page at the NetApp SupportSite.

Related information

Fabric-attached MetroCluster Brocade Switch Configuration Guide - https://now.netapp.com/NOW/knowledge/docs/docs.cgiData ONTAP Information Library - http://now.netapp.com/NOW/knowledge/docs/ontap/ontap_index.shtml

Required toolsLists the tools you need to install the active/active configuration.

The following list specifies the tools you need to install the MetroCluster configuration:






• #1 and #2 Phillips screwdrivers• Hand level• Marker

Required equipmentWhen you receive your MetroCluster, you should receive the equipment listed in the following table.See the System Configuration Guide at the NetApp Support Site to confirm your storage system type,storage capacity, and so on.

Note: For fabric-attached MetroClusters, use the information in the System Configuration Guidelabeled for MetroClusters. For stretch MetroClusters, use the information in the SystemConfiguration Guide labeled “for HA Environments.”

Required equipment Stretch MetroCluster Fabric-attachedMetroCluster

Storage system Two of the same type of storage systems.

Storage See the System Configuration Guide at now.netapp.com.

FibreBridge 6500N (if youare attaching SAS diskshelves)

Two FibreBridge 6500N units are required for each stack of SAS diskshelves.

cluster interconnectadapter

InfiniBand adapter (Required only forsystems that do not use an NVRAM5 orNVRAM6 adapter, which functions asthe cluster interconnect adapter.)

FC-VI adapter (Required only for the31xx dual-controller systems.)

Note: When the FC-VI adapter isinstalled in a31xx system, the internalInfiniBand interconnect isautomatically deactivated.

FC-VI adapter

FC-AL or FC HBA (FCHBA for Disk) adapters

Two or four Fibre Channel HBAs. These HBAs are required for 4-Gbps MetroCluster operation. Onboard ports can be used for 2-Gbpsoperation.

Note: The ports on the Fibre Channel HBAs are labeled 1 and 2.However, the software refers to them as A and B. You see theselabeling conventions in the user interface and system messagesdisplayed on the console.



Required equipment Stretch MetroCluster Fabric-attachedMetroCluster

Fibre Channel switches N/A Two pairs of Brocadeswitches

Note: The FibreChannel switches mustbe of the same type. Amixture of switch types(such as Brocade 300and Brocade 5100switches) is not allowed.

SFP (Small FormPluggable) modules

N/A Two or four long-distancefor inter-switch links,depending on whether youare using dual inter-switchlinks. The type of SFPneeded depends on thedistance between sites.

One short-distance foreach switch port used.

NVRAM adapter mediaconverter

Only if using fiber cabling. N/A

Cables (provided withshipment unless otherwisenoted)

• Four SC/LC (standard connector tolow-profile connector) controller-to-disk shelf cables

• Two SC/LC IB cluster adapter cables• Four SC/LC or LC/LC cables

Note: For information about requiredcables, see the MetroClusterCompatibility Matrix on the NetAppSupport Site.

• LC/LC controller-to-switch cables

• SC/LC (for DS14) orLC/LC (for DS14mk2FC) disk shelf-to-switch cables

• Two LC/LC inter-switch link cables, notprovided in theshipment

• Multiple disk shelf-to-disk shelf cables


MetroCluster and software-based disk ownershipSystems using software-based disk ownership in a MetroCluster require different configuration thansystems using hardware-based disk ownership.

Some systems use software-based disk ownership to control which disks in a disk shelf loop belongto which controller and pool.

• Software commands in Data ONTAP are used to assign disks, or they are auto-assigned by thesoftware.This is because disk ownership is determined by the software, rather than by the physical cablingof the shelves.

• Systems that use software disk ownership require different cabling of their disk shelves when youconfigure your MetroCluster.This is because different Brocade port usage rules are used with software-based disk ownership.

For details about software-based disk ownership, see the Data ONTAP Storage Management Guide.

The 4-Gbps FC-VI adapter requires software disk ownership

If you want to take advantage of the performance provided by the 4-Gbps adapter, you must upgradeto a system that uses software-based disk ownership.

Converting an active/active configuration to a fabric-attached MetroCluster

With the correct hardware, you can reconfigure an active/active configuration to a fabric-attachedMetroCluster.

Before you begin

• If you are upgrading an existing active/active configuration to a MetroCluster configuration, youmust upgrade disk firmware to the latest version. After upgrading disk firmware, you mustpower-cycle the affected disk drives to ensure that they work correctly in a fabric-attachedMetroCluster. You can download the latest disk firmware from support.netapp.com/.

• If you are upgrading from an existing active/active configuration on a system that supports bothsoftware-based and hardware-based disk ownership and is currently using software-based diskownership, you must convert disk assignment to hardware ownership before the upgrade.However, converting an active/active configuration from software-based disk ownership tohardware-based disk ownership is a complicated process. If done incorrectly, your system mightnot boot. You are advised to contact technical support for assistance with this conversion.

• If you are upgrading a31xx system, the resulting upgraded system can only have one controller ineach chassis. If you have a chassis with two controllers, you must move one controller to a new



chassis to form the partner node of the MetroCluster. You must also obtain and install the FC-VIinterconnect card on both systems.

Note: For details about this conversion process, see TR-3517, MetroCluster Upgrade PlanningGuide, on support.netapp.com.

Steps

1. Update Data ONTAP, storage system firmware, and disk firmware, as described in the DataONTAP Upgrade Guide, making sure to shut down the nodes to the boot prompt.

2. Remove any ATA drives in the configuration.

ATA drives are not supported in a MetroCluster configuration.

3. Move the NVRAM adapter and FC-VI adapter to the correct slots for your model, as shown bythe System Configuration Guide at the NetApp Support Site.

4. Determine your switch and general configuration by completing the planning worksheet.

5. Set up and configure the local switches, and verify your switch licenses, as described in theBrocade Switch Configuration Guide for Fabric-attached MetroClusters.

You can find this document on the Brocade Switch Description Page on the NetApp Support Site.

Note: The configuration and firmware requirements for Brocade switches in a MetroClusterenvironment are different from the requirements for switches used in SAN environments.Always refer to MetroCluster documentation when installing and configuring yourMetroCluster switches:

• The MetroCluster Compatibility Matrix• The Brocade Switch Description Page• The Brocade Switch Configuration Guide for Fabric-attached MetroClusters

6. Cable the local node.

7. Install the Data ONTAP licenses in the following order:

a. cluster

b. syncmirror_local

c. cluster _remote

8. Configure the local node depending on the type of active/active configuration:

If you are converting a... Then...

Standard active/active configuration Set up mirroring and configure the local node.

Stretch MetroCluster Configure the local node.

9. Transport the partner node, disk shelves, and switches to the remote location.



10. Set up the remote node, disk shelves, and switches.

After you finish

Configure the MetroCluster.

Related concepts

Configuring an active/active configuration on page 99


Related tasks

Cabling Node A on page 76

Cabling Node B on page 82

Disabling the change_fsid option in MetroCluster configurations on page 104

Planning the fabric-attached MetroCluster installation on page 73

Upgrading an existing MetroClusterYou can upgrade an existing MetroCluster on a system using hardware-based disk ownership to aMetroCluster on a system using software-based disk ownership (FAS3040, FAS3070, or FAS60xxsystems). This is useful when you are upgrading to 4-Gbps cluster interconnect support, whichrequires software-based disk ownership.

About this task

When using the typical hardware upgrade procedure you upgrade your software on the old systemand then use the disk upgrade_ownership command to apply software-based ownership to thedisks. You then perform the hardware upgrade.

In the following procedure, you perform the hardware upgrade prior to using the diskupgrade_ownership command. This is because the old system hardware does not support the newfeatures of the disk upgrade_ownership command. For the Data ONTAP 7.2.3 (or later) versionof the disk upgrade_ownership command to run successfully, you must issue it on a system thatsupports software-based disk ownership.

Steps

1. Halt the system, and then turn off the controller and disk shelves.

2. Remove the existing controller from the rack or system cabinet and install the FAS3040,FAS3070, or FAS60xx system in its place.


When replacing the controllers, use the same cabling to the Brocade switches and the disk shelvesas the original controller. For the upgrade to work, you must retain the original cabling until yourun the disk upgrade_ownership command later in this procedure.

3. Power on the disk shelves.

4. To reassign disk ownership to software-based disk ownership, complete the following substepson both controllers:

a. Power on the system and boot the system into Maintenance mode.

For more information, see the Data ONTAP System Administration Guide.

b. Enter the following command at the firmware prompt:

disk upgrade_ownership

This command converts the system to software-based disk ownership. Data ONTAP assignsall the disks to the same system and pool that they were assigned to for the hardware-baseddisk ownership.

See the Data ONTAP Storage Management Guide for detailed information about software-based disk ownership.

5. Verify disk ownership information by entering the following command:

disk show -v

Disk assignment is now complete.

6. Clear the mailboxes by entering the following commands:

mailbox destroy local

mailbox destroy partner

7. Enter the following command to exit Maintenance mode:

halt

8. Enter the following command for each required license:

Example

license add xxxxxx

xxxxx is the license code you received for the feature.

9. Enter the following command to reboot the node:

reboot

10. Configure the RLM, if applicable, as described in the Data ONTAP Software Setup Guide.

11. Recable the connections to the Brocade switches to conform to the virtual channel rules for theswitch.


After you finish

Configure the MetroCluster.

Related concepts



Switch bank rules and virtual channel rules on page 75

Related tasks

Cabling Node A on page 76

Cabling Node B on page 82


Cabling a stretch MetroClusterThe process to cable a stretch MetroCluster is the same as a mirrored active/active configuration.However, your systems must meet the requirements for a stretch MetroCluster.

Related concepts


Setup requirements and restrictions for stretch MetroCluster configurations on page 27


Related tasks

Cabling a mirrored active/active configuration on page 47

Cabling a stretch MetroCluster between single-enclosureactive/active configuration systems

If you are configuring a stretch MetroCluster between single-enclosure active/active configurationsystems (for example, 31xx systems), you must configure FC-VI interconnect adapter connectionsbetween the controllers.

About this task

Some storage systems support two controllers in the same chassis. You can configure two dual-controller systems into a pair of MetroClusters. In such a configuration, the internal InfiniBandconnections between the controllers are automatically deactivated. Therefore, the two controllers inthe chassis are no longer in an active/active configuration with each other. Each controller is


connected through FC-VI connections to another controller of the same type, so that the fourcontrollers form two independent MetroClusters configuration.

Steps

1. Connect port A of the FC-VI adapter on the top controller of the local site to port A of thecorresponding FC-VI adapter at the remote site.

2. Connect port B of the FC-VI adapter on the top controller of the local site to port B of thecorresponding FC-VI adapter at the remote site.

3. Repeat steps 1 and 2 for connecting the FC-VI adapter on the bottom controller.

4. Cable the disk shelf loops for the stretch MetroCluster formed by the top controllers as describedin the procedure for cabling a mirrored active/active configuration.

5. Cable the disk shelf loops for the stretch MetroCluster formed by the bottom controllers asdescribed in the procedure for cabling a mirrored active/active configuration.

Related concepts

Stretch MetroCluster configuration on single enclosure active/active configurations on page 26

Related tasks

Cabling a mirrored active/active configuration on page 47

Changing the default configuration speed of a stretchMetroCluster

The distance between your nodes and the FC-VI adapter speed dictates the default configurationspeed of your stretch MetroCluster. If the distance between nodes is greater than the supporteddefault configuration speed, you must change the default configuration speed.

Before you begin

The stretch MetroCluster default configuration speed must conform to the Stretch MetroCluster setuprequirements and restrictions.

About this task

• You enter the commands at the boot environment prompt, which can be CFE> or LOADER>,depending on your storage system model.

• You must perform these steps at both the nodes if they are configured at different speeds.


Steps

1. At the storage console prompt, halt the storage system by entering the following command:

halt

2. Reset the configuration speed by entering the following commands.

If you want to setthe speed to...

Then...

4 Gb a. Enter the following command:

setenv ispfcvi-force-4G-only? True

b. If you previously modified the speed to 2 Gb, ensure that the 2-Gb port speed isnot set by entering the following command:

unsetenv ispfcvi-force-2G-only?

c. Verify that your system is unconfigured for 2 Gb by entering the followingcommand:

printenv ispfcvi-force-2G-only?

The storage system console displays output similar to the following:

Variable Name Value-------------------- --------------ispfcvi-force-2G-only? *** Undefined ***

d. Verify that your storage system is configured for 4 Gb by entering the followingcommand:



Variable Name Value-------------------- --------------ispfcvi-force-4G-only? true


If you want to setthe speed to...

Then...


setenv ispfcvi-force-2G-only? True

b. If you previously modified the default speed to 4 Gb, ensure that the 4-Gb speedis not set by entering the following command:


c. Verify that your storage system is unconfigured for 4 Gb by entering thefollowing command:



Variable Name Value-------------------- --------------------------------------------------ispfcvi-force-4G-only? *** Undefined ***

d. Verify that your storage system is configured for 2 Gb by entering the followingcommand:


If your storage system is configured correctly, the system console displays outputsimilar to the following:

Variable Name Value-------------------- --------------------------------------------------ispfcvi-force-2G-only? true

3. Boot the storage system by entering the following command:

boot_ontap

Resetting a stretch MetroCluster configuration to the defaultspeed

If you modified the default configuration speed in a stretch MetroCluster using an FC-VI adapter,you can reset the speed to the default configuration speed by using the unsetenv command at theboot environment prompt.

About this task

• The boot environment prompt can be CFE> or LOADER>, depending on your storage systemmodel.


• The steps require you to unset the previously configured speed, but because the speed is set todefault automatically, you do not need to set the default speed explicitly.

Steps

1. At the storage prompt, halt the system by entering the following command:

halt

2. Reset the configuration speed.

If you want toreset thespeed from...

Then...



b. Verify that your system is unconfigured for 4 Gb by entering the following command:


The system console displays output similar to the following:

Variable Name Value-----------------------------ispfcvi-force-4G-only? *** Undefined ***



b. Verify that your system is unconfigured for 2 Gb by entering the following command:


3. Boot the storage system by entering the following command:

boot_ontap

Cabling a fabric-attached MetroClusterYou cable the fabric-attached MetroCluster so that the controller and the disk shelves at each site areconnected to Brocade switches. In turn, the Brocade switches at one site are connected through inter-switch links to the Brocade switches at the other site.

Before you begin

To cable a fabric-attached MetroCluster, you must be familiar with active/active configurations, theBrocade command-line interface, and synchronous mirroring. You must also be familiar with thecharacteristics of fabric-attached MetroClusters. You must also have the following information:


• Correct Brocade licenses for each switch• Unique domain IDs for each of the switches

Note: You can use the switch numbers (1, 2, 3, and 4) as the switch Domain ID.

• Ethernet IP address for both the switches and nodes

Note: The switches ship with a default IP address (10.77.77.77), which you can use if theswitches are not attached to a network.

• Ethernet subnetmask• Gateway address

About this task

A fabric-attached MetroCluster involves two nodes at physically separated sites. To differentiatethese nodes in this documentation, the guide refers to the two nodes as Node A and Node B.

Complete the following tasks in the order shown:

Steps

1. Planning the fabric-attached MetroCluster installation on page 73

2. Configuration differences for fabric-attached MetroClusters on single-enclosure active/activeconfiguration on page 74

3. Configuring the switches on page 75

4. Cabling Node A on page 76

5. Cabling Node B on page 82

6. Assigning disk pools (if you have software-based disk ownership) on page 87

7. Verifying disk paths if you have software-based disk ownership on page 88

Related concepts

Setup requirements and restrictions for fabric-attached MetroClusters on page 31




Related tasks


Planning the fabric-attached MetroCluster installationYou must fill out the fabric-attached MetroCluster worksheet to record specific cabling informationabout your fabric-attached MetroCluster. You must identify several pieces of information that youuse during configuration procedures. Recording this information can reduce configuration errors.

Step

1. Fill in the following tables.

Each site has two Brocade Fibre Channel switches. Use the following table to record theconfigured names, IP addresses, and domain IDs of these switches.

Switchnumber...

At site... Is named... IP address... Domain ID...

1 A

2 A

3 B

4 B

In addition to on-board ports, each site has a FC-VI adapter and two Fibre Channel HBAs thatconnect the node to the switches. Use the following table to record which switch port theseadapters are connected to.

This adapter... At site... Port 1 of this adapteris...

Port 2 of this adapteris...

Cabled toswitch...

Switchport...

Cabled toswitch...

Switchport...

FC-VI adapter A 1 2

B 3 4

FC HBA 1 A 1 2

B 3 4


This adapter... At site... Port 1 of this adapteris...

Port 2 of this adapteris...

Cabled toswitch...

Switchport...

Cabled toswitch...

Switchport...

FC HBA 2 A 1 2

B 3 4

Disk shelves at each site connect to the Fibre Channel switches. Use the following table to recordwhich switch port the disk shelves are connected to.

Disk shelf... At site... Belonging to... Connects toswitches...

On switch port...

1 A Node A Pool 0 1 and 2

2

3 Node B Pool 1

4

5 B Node B Pool 0 3 and 4

6

7 Node A Pool 1

8

Configuration differences for fabric-attached MetroClusters on single-enclosure active/active configuration

When configuring a fabric-attached MetroCluster between single-enclosure active/activeconfiguration (systems with two controllers in the same chassis), you get two separate MetroClusterconfigurations.

A single-enclosure active/active configuration can be connected to another active/activeconfiguration to create two separate fabric-attached MetroClusters. The internal InfiniBandconnection in each system is automatically deactivated when the FCVI card is installed in thecontroller.

You must cable each fabric-attached MetroCluster separately by using the normal procedures foreach and assign the storage appropriately.

Related concepts

Fabric-attached MetroCluster configuration on single-enclosure active/activeconfigurations on page 30


Configuring the switchesTo configure the switches, you refer to the Brocade Switch Configuration Guide for Fabric-attachedMetroClusters for your Brocade switch model. The Brocade switch configuration for a MetroClusteris different than the one used for a SAN configuration.

Step

1. To configure your Brocade switches, see the Brocade Switch Configuration Guide for Fabric-attached MetroClusters for your switch model. You can find this document on the MetroClusterSwitch Description Page at the NetApp Support Site at support.netapp.com. Scroll down the pageto Product Documentation > Configuration Guides.

Note: The configuration and firmware requirements for Brocade switches in a MetroClusterenvironment are different from the requirements for switches used in SAN environments.Always refer to MetroCluster documentation, such as the MetroCluster Compatibility Matrixor the MetroCluster Switch Description Page, when installing and configuring yourMetroCluster switches.

After you finish

Proceed to configure Node A.

Related concepts

Switch bank rules and virtual channel rules on page 75

Switch bank rules and virtual channel rules

You must follow the correct switch bank rules or virtual channel rules on the Brocade switches.

If your system uses hardware-based disk ownership, you must use the switch bank rules whencabling the Brocade switch. This ensures that switch traffic is distributed across the switch quadrantsto reduce potential bottlenecks.

Figure 1: Brocade switch showing which ports belong to which switch banks andpools

If your system does not use hardware-based disk ownership, use the switch virtual channel (VC)rules when cabling the switch. In this case, switch traffic is distributed across VCs to avoid



bottlenecks. The FC-VI and inter-switch links are cabled to ports in one VC, and the disk shelf andcontroller connections are cabled to ports in another VC.

Virtual channel Ports

2 0, 4, 8, 12, 16, 20, 32, 36

3 1, 5, 9, 13, 17, 21, 33, 37

4 2, 6, 10, 14, 18, 22, 26, 30, 34, 38

5 3, 7, 11, 15, 19, 23, 27, 31, 35, 39

Related information

Brocade Switch Configuration Guide for Fabric MetroCluster - http://now.netapp.com/NOW/download/software/sanswitch/fcp/Brocade/mc_ontap641_fabric_index.shtml

Cabling Node ATo cable the local node (Node A), you need to attach the controller and the disk shelves to theswitches, connect the cluster interconnect to the switches, and ensure that the disk shelves in theconfiguration belong to the correct pools.

About this task


Steps

1. Cabling the controller when you have software-based disk ownership on page 76

2. Cabling the shelves when you have software-based disk ownership on page 78

3. Cabling the FC-VI adapter and inter-switch link when you have software-based diskownership on page 80

Cabling the controller when you have software-based disk ownership

You can use this procedure to cable the Fibre Channel ports on the controller to the Brocade switcheswhen your system uses software-based disk ownership.

Before you begin

• Select one virtual channel on the switch for the cluster interconnect connections. The followingexamples use virtual channel 2, which includes ports 0, 4, 8, and 12.


http://now.netapp.com/NOW/download/software/sanswitch/fcp/Brocade/mc_ontap641_fabric_index.shtml

http://now.netapp.com/NOW/download/software/sanswitch/fcp/Brocade/mc_ontap641_fabric_index.shtml

• If you are using a dual-port HBA, connecting both ports of the HBA to the same switch portnumber can make it easier to cable and administer your MetroCluster. (However, this is notrequired.)For example, if port 1 of the HBA is connected to port 1 of Switch 1, you should connect port 2of that HBA to port 1 of Switch 2.

• Both Fibre Channel ports on the same dual-port HBA (or adjacent pairs of onboard ports) shouldnever be connected to the same switch. You must connect one port to one switch and the otherport to the other switch.For example, if onboard port 0a is connected to Switch 3, you should not connect onboard port 0bto Switch 3; you must connect port 0b to Switch 4.

About this task

Steps

1. Determine which Fibre Channel ports on your system that you want to use and create a listshowing the order you want to use them.

Note: The numbers in the example refer to the preferred order of usage, not the port ID. Forexample, Fibre Channel port 1 might be port e0a on the controller.

2. Cable the first two Fibre Channel ports of Node A to the same numbered ports on Switch 1 andSwitch 2. For example, port 1.

They must not go to ports in the virtual channel that you have reserved for the FC-VI and inter-switch link connections. In the example, we are using virtual channel 2 for the FC-VI and inter-switch link. Virtual channel 2 includes ports 0, 4, 8, and 12.

3. Cable the second two Fibre Channel ports of Node A to the same numbered ports on Switch 1 andSwitch 2. For example, port 2.

Again, they must not go to ports in the virtual channel that you have reserved for the FC-VI andinter-switch link connections. In the example, ports 0, 4, 8, and 12 are excluded.

Note: The switches in the example are 16-port switches.


After you finish

Proceed to cable disk shelves to the switches.

Related concepts

Determining which Fibre Channel ports to use for Fibre Channel disk shelf connections on page 41

Cabling the shelves when you have software-based disk ownership

You must cable the disk shelf loops on Node A directly to the Brocade switches.

Before you begin

• You can connect the disk shelves to any ports on that are not on the virtual channel reserved forthe FC-VI adapter and the inter-switch link.

• Both disk shelf modules on the same loop must be connected to the same switch port number.For example, if the A Channel of the first loop for the local node’s disks is connected to Switch 1,port 8, then the B Channel for that loop must be connected to Switch 2, port 8.

• Both switches at a site must be the same model and have the same number of licensed ports.

About this task


Note: You can cable a maximum of two disk shelves on each loop.

Steps

1. Connect the Node A pool 0 disk shelves to the switches by completing the following substeps:

a. Connect the Input port of the A module on disk shelf 1 to any available port on Switch 2 otherthan ports 0, 4, 8, and 12. In the example, switch port 3 is used.

b. Connect the Input port of the B module on disk shelf 1 to the same port on Switch 1. Theexample uses switch port 3.

c. Connect disk shelf 1 to disk shelf 2 by connecting the Output ports of the module of disk shelf1 to the Input ports of the corresponding module of the next disk shelf.

d. If your disk shelf modules have terminate switches, set them to Off on all but the last diskshelf in the disk pool, then set the terminate switches on the last disk shelf to On.


Note: ESH2 and ESH4 modules are self-terminating and therefore do not have a terminateswitch.

2. Connect the Node B pool 1 disk shelves to the switches by completing the following substeps:

a. Connect the Input port of the module Channel A on disk shelf 3 to any available port onSwitch 2 other than ports 0, 4, 8, and 12. The example uses switch port 5.

b. Connect the Input port of the module Channel B on disk shelf 3 to the same port on Switch 1.The example uses switch port 5.



3. If you have more than one loop, connect the other loops in the same manner.

After you finish

Proceed to cable the FC-VI adapter and inter-switch connections.

Cabling the FC-VI adapter and inter-switch link when you have software-based diskownership

Describes how to cable the cluster interconnect and inter-switch link on Node A.

Before you begin

Each port on the Interconnect (IC) cards must be connected to the same fabric.

For example, if Port A of the IC card on the local node is connected to Switch 1, and Port A of the ICcard on the remote node is connected to Switch 3, then Switch 1 and Switch 3 must be connected bythe inter-switch link, thereby connecting them to the same fabric.


About this task

Steps

1. Using the ports in the virtual channel you have selected for the FC-VI and inter-switch linkconnections, connect one port of the FC-VI adapter on switch 1 and the second port to the sameport on switch 2.

In the example we are using virtual channel 2, including ports 0, 4, 8, and 12, for the FC-VI andinter-switch link connections.

Note: There should be one FC-VI adapter connection for each switch. Make sure that you havethe FC-VI adapter in the correct slot for your system, as shown in the System ConfigurationGuide at http://now.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml.

2. Connect an inter-switch link cable to a port in the selected virtual channel on each switch, or, ifusing a dual inter-switch link, connect two cables in the selected virtual channel.

In the example we are using virtual channel 2, which includes ports 0, 4, 8, and 12, and are usingports 8 and 12 on switch 1 and switch 2 for the inter-switch links.

Note: If using dual inter-switch links, traffic isolation must be configured on the switches.

After you finish

Proceed to cable Node B.


Cabling Node BTo cable the remote node (Node B), you need to attach the controller and the disk shelves to theswitches, connect the cluster interconnect to the switches, and ensure that the disk shelves in theconfiguration belong to the correct pools.

About this task


Steps

1. Cabling the controller when you have software-based disk ownership on page 82

2. Cabling the shelves when you have software-based disk ownership on page 84

3. Cabling the FC-VI adapter and inter-switch link when you have software-based diskownership on page 85

Cabling the controller when you have software-based disk ownership

You can use this procedure to cable the Fibre Channel ports on the controller to the Brocade switcheswhen your system uses software-based disk ownership.

Before you begin

• Select one virtual channel on the switch for the cluster interconnect connections. The followingexamples use virtual channel 2, which includes ports 0, 4, 8, and 12.

• If you are using a dual-port HBA, connecting both ports of the HBA to the same switch portnumber can make it easier to cable and administer your MetroCluster. (However, this is notrequired.)For example, if port 1 of the HBA is connected to port 1 of Switch 1, you should connect port 2of that HBA to port 1 of Switch 2.

• Both Fibre Channel ports on the same dual-port HBA (or adjacent pairs of onboard ports) shouldnever be connected to the same switch. You must connect one port to one switch and the otherport to the other switch.For example, if onboard port 0a is connected to Switch 3, you should not connect onboard port 0bto Switch 3; you must connect port 0b to Switch 4.


About this task

Steps

1. Determine which Fibre Channel ports on your system that you want to use and create a listshowing the order you want to use them.

Note: The numbers in the example refer to the preferred order of usage, not the port ID. Forexample, Fibre Channel port 1 might be port e0a on the controller.

2. Cable the first two Fibre Channel ports of Node B to the same numbered ports Switch 3 andSwitch 4. For example, port 1.

They must go to ports in the virtual channel that you have reserved for the FC-VI and inter-switchlink connections. In the example, we are using virtual channel 2 for the FC-VI and inter-switchlink. Virtual channel 2 includes ports 0, 4, 8, and 12.

3. Cable the second two Fibre Channel ports of Node B to the same numbered ports Switch 3 andSwitch 4. For example, port 2.

Again, they must not go to ports in the virtual channel that you have reserved for the FC-VI andinter-switch link connections. In the example, ports 0, 4, 8, and 12 are excluded.

After you finish

Proceed to cable disk shelves to the switches.

Related concepts

Determining which Fibre Channel ports to use for Fibre Channel disk shelf connections on page 41


Cabling the shelves when you have software-based disk ownership

You must cable the disk shelf loops on Node B directly to the Brocade switches.

Before you begin

• You can connect the disk shelves to any ports on that are not on the virtual channel reserved forthe FC-VI adapter and the inter-switch link.

• Both disk shelf modules on the same loop must be connected to the same switch port number.For example, if the A Channel of the first loop for the local node’s disks is connected to Switch 1,port 8, then the B Channel for that loop must be connected to Switch 2, port 8.

• Both switches at a site must be the same model and have the same number of licensed ports.

About this task


Note: You can cable a maximum of two disk shelves on each loop.

Steps


a. Connect the Input port of the A module on disk shelf 5 to any available port on Switch 4 thatis not in the virtual channel reserved for the FC-VI and inter-switch link connections. Theexample uses switch port 3.

b. Connect the Input port of the B module on disk shelf 5 to the same port on Switch 3. Theexample uses switch port 3.



Note: ESH2 and ESH4 modules are self-terminating and therefore do not have a terminateswitch.


a. Connect the Input port of the module Channel A on disk shelf 7 to any available port onSwitch 4 that is not in the virtual channel reserved for the FC-VI and inter-switch linkconnections. The example uses switch port 5.

b. Connect the Input port of the module Channel B on disk shelf 7 to the same port on Switch 3.The example uses switch port 5.



3. If you have more than one loop, connect the other loops in the same manner.

After you finish

Proceed to cable the FC-VI adapter and inter-switch connections.

Cabling the FC-VI adapter and inter-switch link when you have software-based diskownership

You must cable the cluster interconnect and inter-switch link on Node B.

Before you begin

Each port on the Interconnect (IC) cards must be connected to the same fabric.


For example, if port A of the IC card on the local node is connected to switch 1, and port A of the ICcard on the remote node is connected to switch 3, then switch 1 and switch 3 must be connected bythe inter-switch link, thereby connecting them to the same fabric.

About this task

Steps

1. Connect one port of the FC-VI adapter to a port in the virtual channel that you have reserved forthe FC-VI and inter-switch link connections.

In the example, port 0 on switch 1 and port 0 on switch 2 is used.

Note: There should be one FC-VI adapter connection for each switch. Make sure that you havethe FC-VI adapter in the correct slot for your system, as shown in the System ConfigurationGuide at http://now.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml.

2. Connect an inter-switch link cable to a port in the selected virtual channel on each switch, or ifusing dual inter-switch links, connect two cables in the selected virtual channel.

In the example we are using virtual channel 2, which includes ports 0, 4, 8, and 12, and are usingport 8 and port 12 on switch 1 and switch 2 for the inter-switch links.

Note: If using dual inter-switch links, traffic isolation must be configured on the switches.

After you finish

Proceed to assign disks to disk pools.

Related tasks

Assigning disk pools (if you have software-based disk ownership) on page 87


Assigning disk pools (if you have software-based disk ownership)If your system uses software-based disk ownership, you must assign the attached disk shelves to theappropriate pools.

About this task

You can explicitly assign disks on the attached disk shelves to the appropriate pool with the diskassign command. Using wildcards in the command enables you to assign all the disks on a diskshelf with one command.

The following table shows the pool assignments for the disk shelves in the example used in thissection.

Disk shelf... At site... Belongs to... And is assigned tothat node's...

Disk shelf 1 Site A Node A Pool 0

Disk shelf 2

Disk shelf 3 Node B Pool 1

Disk shelf 4

Disk shelf 5 Site B Node B Pool 0

Disk shelf 6

Disk shelf 7 Node A Pool 1

Disk shelf 8

Note: Pool 0 always contains the disks that are local to (at the same site as) the storage system thatowns them.

Pool 1 always contains the disks that are remote to the storage system that owns them.

Steps

1. Boot Node A into Maintenance mode, if you haven’t already.

2. Assign the local disks to Node A pool 0 by entering the following command at the console:

disk assign switch2:port3.* -p0

This indicates that the disks attached to port 3 of switch 2 are assigned to pool 0. The asterisk (*)indicates that all disks attached to the port are assigned.

3. Assign the remote disks to Node A pool 1 by entering the following command at the console:




4. Boot Node B into Maintenance mode, if you haven’t already.

5. Assign the local disks to Node B pool 0 by entering the following command at the console:



6. Assign the remote disks to Node B pool 1 by entering the following command at the console:



After you finish

Proceed to verify the disk paths on the system.

Verifying disk paths if you have software-based disk ownershipUse this procedure to verify your disk paths if you have software-based disk ownership.

Steps

1. Boot Node A into normal mode, if necessary.

2. Enter the following command to confirm that your aggregates and volumes are operational andmirrored:

aggr status

See the Data ONTAP Storage Management Guide for information on the aggr statuscommand.

3. Repeat steps 1 and 2 on Node B.

Required connections for using uninterruptible powersupplies with MetroCluster configurations

You can use a UPS (Uninterruptible Power Supply) with your MetroCluster. The UPS enables thesystem to fail over gracefully if power fails for one of the nodes, or to shut down gracefully if powerfails for both nodes. You must ensure that the correct equipment is connected to the UPS.

The equipment that you need to connect to the UPS depends on how widespread a power outage youwant to protect against. Always connect both controllers, any Fibre Channel switches in use, and anyinter-switch link infrastructure (for example, a Dense Wavelength Division Multiplexing, orDWDM) to the UPS.

You can leave the disks on the regular power supply. In this case, if power is interrupted to one site,the controller can access the other plex until it shuts down or power is restored. If, however, power is


interrupted to both sites at the same time and the disks are not connected to the UPS, theMetroCluster cannot shut down gracefully.


Reconfiguring an active/active configuration intotwo stand-alone systems

To divide an active/active configuration so that the nodes become stand-alone systems withoutredundancy, you must disable the active/active software features and then remove the hardwareconnections.

About this task

This procedure applies to all active/active configurations regardless of disk shelf type.

Steps

1. Ensuring uniform disk ownership within disk shelves and loops in the system on page 91

2. Disabling the active/active software on page 92

3. Reconfiguring nodes using disk shelves for stand-alone operation on page 93

4. Requirements when changing a node using array LUNs to stand-alone on page 95

5. Reconfiguring nodes using array LUNs for stand-alone operation on page 96

Ensuring uniform disk ownership within disk shelves andloops in the system

In systems using software-based disk ownership, if a disk shelf or loop contains a mix of disksowned by Node A and Node B, you must use this procedure to move the data and make diskownership uniform within the disk shelf or loop.

Before you begin

You must ensure the following:

• Disk ownership is uniform within all disk shelves and loops in the system• All the disks within a disk shelf or loop belong to a single node and pool

About this task

Note: It is a best practice to always assign all disks on the same loop to the same node and pool.

Steps

1. Use the following command to identify any disk shelves or loops that contain both disksbelonging to Node A and disks belonging to Node B:

Reconfiguring an active/active configuration into two stand-alone systems | 91

disk show -v

2. Determine which node the disk shelf or loop with mixed ownership will be attached to when theactive/active feature is unconfigured and record this information.

For example, if the majority of the disks in the loop belong to Node A, you probably want theentire loop to belong to stand-alone Node A.

After you finish

Proceed to disable the active/active software.

Disabling the active/active softwareYou need to disable the active/active configuration in the software before reconfiguring the hardwareto completely unconfigure the active/active feature.

Before you begin

Before performing this procedure you must ensure that all loops and disk shelves in the systemcontain disks belonging to one or the other nodes. The disk shelves and loops can't contain a mix ofdisks belonging to Node A and Node B. In any disk shelves or loops containing such a mix of disks,you must move data.

Steps

1. Enter the following command on either node console:

cf disable

2. Disable the cluster license by entering the following command:

license delete cluster

3. Open the /etc/rc file with a text editor and remove references to the partner node in the ifconfigentries, as shown in the following example:

Example

Original entry:

ifconfig e0 199.9.204.254 partner 199.9.204.255

Edited entry:

ifconfig e0 199.9.204.254

4. Repeat Step 1 through Step 3 on the partner node.

After you finish

Proceed to reconfigure the hardware.


Reconfiguring nodes using disk shelves for stand-aloneoperation

You can use this procedure to reconfigure the hardware if you want to return to a single-controllerconfiguration.

Before you begin

You must disable the active/active software.

Steps

1. Halt both nodes by entering the following command on each console:

halt

2. Using the information you recorded earlier, in the disk shelves or loops with mixed storage,physically move the disks to a disk shelf in a loop belonging to the node that owns the disk. Forexample, if the disk is owned by Node B, move it to a disk shelf in a loop that is owned by NodeB.

Note: Alternatively, you can move the data on the disks using a product such as Snapshotsoftware, rather than physically moving the disk. See the Data ONTAP Data Protection OnlineBackup and Recovery Guide.

After moving the data from the disk you can zero the disk and use the diskremove_ownership command to erase the ownership information from the disk. See the DataONTAP Storage Management Guide.

3. If you are completely removing one node, so that all the disk shelves will belong to a singlestand-alone node, complete the following substeps:

a. Boot the node being removed into Maintenance mode, as described in the Data ONTAPSystem Administration Guide.

b. Use the disk reassign command and reassign all disk shelves so that they all belong to thenode that remains.

The disk reassign command has the following syntax:

disk reassign [-o <old_name> | -s <old_sysid>] [-n <new_name>] -d<new_sysid>

c. Halt the node by entering the following command:

halt

4. Turn off the power to each node, then turn off the power to the disk shelves and unplug themfrom the power source.


5. Ground yourself, then remove the cluster interconnect cables from both nodes. See the hardwaredocumentation for your system for more details.

6. Move or remove the adapter used for the cluster interconnect:

If your system uses a... Then...

cluster interconnect adapter or an FC-VIadapter

Remove the adapter from the system.

NVRAM5 or NVRAM6 adapter You might need to change the slot position of the adapter.See the System Configuration Guide for details aboutexpansion slot usage for the adapter.

7. Recable the system, depending on the type of system:

If you are converting a... Then...

System with nonmirrored disks a. Disconnect all cabling from the Channel B loop on the local node.

b. Repeat for the partner node.

System with mirrored disks or aredundant Channel B loop

a. Connect the local node to the open Channel B loop in its localdisk shelves, as described in the appropriate disk shelf guide.

b. Repeat for the partner node.

8. Power on the disk shelves, then the individual nodes, monitoring the system console for errormessages during the boot process.

9. Run all system diagnostics at the boot prompt by entering the following command on the systemconsole:

boot diags

10. Unset the partner system ID by entering the following command at the prompt:

unsetenv partner-sysid

11. Boot the node by entering the following command:

boot

12. Check active/active configuration status by entering the following command:

cf status

If the active/active configuration is disabled, you see the following output:Failover monitor not initialized

13. Repeat Step 1 through Step 10 for the partner node.

Related concepts

Requirements when changing a node using array LUNs to stand-alone on page 95


Related tasks

Reconfiguring nodes using array LUNs for stand-alone operation on page 96

Requirements when changing a node using array LUNs tostand-alone

After uncoupling V-Series systems in an active/active configuration, you might need to performadditional reconfiguration related to Data ONTAP ownership of array LUNs.

The following table summarizes the requirements when uncoupling a storage system using arrayLUNs from an active/active configuration.

If you want to... Requirements foruncoupling systems are...

Requirements for array LUNassignments to systems are...

Make both systems in thepair stand-alone systems

Remove the active/activeconfiguration softwareand interconnect cabling

No Data ONTAP reconfiguration of arrayLUNs is necessary. Each system cancontinue to own the array LUNs assignedto it.

Remove one system in thepair from service

Remove the active/activeconfiguration softwareand interconnect cabling

After uncoupling the pair, you must doone of the following:

• If you want to continue to use thearray LUNs for Data ONTAP,reassign the array LUNs owned by thesystem you are removing to anotherstorage system.

• Prepare the array LUNs assigned tothe system you are removing for useby systems that do not run DataONTAP.

Related tasks

Disabling the active/active software on page 92

Reconfiguring nodes using array LUNs for stand-alone operation on page 96


Reconfiguring nodes using array LUNs for stand-aloneoperation

After uncoupling the nodes in active/active configuration, each node can continue to own its assignedarray LUNs, you can reassign its array LUNs to another V-Series system, or you can release thepersistent reservations on the array LUNs so the LUNs can be used by a non Data ONTAP host.

Before you begin

You must disable the active/active configuration software.

About this task

If you want both systems in the active/active configuration to remain in service and operate as stand-alone systems, each system can continue to own the array LUNs that were assigned to it. Eachsystem, as a stand-alone, will continue to see the array LUNs owned by the other system becauseboth systems are still part of the same V-Series neighborhood. However, only the system that is theowner of the array LUNs can read from or write to the array LUN, and the systems can no longer failover to each other.

Steps

1. On each node, halt the node by entering the following command at the console:

halt

2. Turn off the power to each node.

3. Ground yourself, then remove the cluster interconnect cables from both nodes. See the hardwaredocumentation for your system for more details.

4. Move or remove the adapter used for the cluster interconnect.

If your system uses a... Then...

cluster interconnect adapter or an FC-VIadapter

Remove the adapter from the system.

NVRAM5 or NVRAM6 adapter You might need to change the slot position of the adapter.See the System Configuration Guide for details aboutexpansion slot usage for the adapter.

5. On each node, perform the following steps:

a. Power on the node, monitoring the system console for error messages during the boot process.

b. Unset the partner system ID by entering the following command at the prompt:

unsetenv partner-sysid


6. Perform the appropriate step in the following table for what you intend to do with your systemand its storage.

If you want to... Then...

Keep both systems in service asstand-alone systems andcontinue with both systemsowning the array LUNs thatwere already assigned to them

Boot both systems by entering the following command on each system:

boot

Remove one of the systems fromservice but still use the storagethat was assigned to that systemfor Data ONTAP

a. Boot the node being removed into Maintenance mode, as describedin the Data ONTAP System Administration Guide.

b. Use the disk reassign command to reassign all the arrayLUNs so that they all belong to the node that remains.The disk reassign command has the following syntax:

disk reassign [-o <old_name> | -s<old_sysid>] [-n <new_name>] -d <new_sysid>

c. Remove the node from service.

d. Boot the node you are keeping in service by entering the followingcommand:

boot

Remove one of the systems fromservice and use the array LUNsthat are currently assigned to itfor a host that does not run DataONTAP

Release the persistent reservations that Data ONTAP placed on thosearray LUNs so that the storage administrator can use those LUNs forother hosts.

See the Data ONTAP Storage Management Guide for information aboutwhat you need to do to prepare for taking a system using array LUNsout of service.

Related tasks

Disabling the active/active software on page 92


Configuring an active/active configuration

Bringing up and configuring a standard or mirrored active/active configuration for the first time canrequire enabling licenses, setting options, configuring networking, and testing the configuration.

These tasks apply to all active/active configurations regardless of disk shelf type.

Steps

1. Bringing up the active/active configuration on page 99

2. Enabling licenses on page 102

3. Setting options and parameters on page 103

4. Configuration of network interfaces on page 109

5. Downloading and running the HA Configuration Checker utility on page 119

6. Testing takeover and giveback on page 119

Bringing up the active/active configurationThe first time you bring up the active/active configuration, you must ensure that the nodes arecorrectly connected and powered up, and then use the setup program to configure the systems.

Considerations for active/active configuration setupWhen the setup program runs on a storage system in an active/active configuration, it prompts you toanswer some questions specific for active/active configurations.

The following list outlines some of the questions about your installation that you should think aboutbefore proceeding through the setup program:

• Do you want to configure virtual interfaces (VIFs) for your network interfaces?For information about VIFs, see the Data ONTAP Network Management Guide.

Note: You are advised to use VIFs with active/active configurations to reduce SPOFs (single-points-of-failure).

• How do you want to configure your interfaces for takeover?

Note: If you do not want to configure your network for use in an active/active configuration whenyou run setup for the first time, you can configure it later. You can do so either by running setupagain, or by using the ifconfig command and editing the /etc/rc file manually. However, youmust provide at least one local IP address to exit setup.

Related concepts

Configuration of network interfaces on page 109

Configuring an active/active configuration | 99

Related tasks

Configuring shared interfaces with setup on page 100

Configuring dedicated interfaces with setup on page 101

Configuring standby interfaces with setup on page 101

Configuring shared interfaces with setupDuring setup of the storage system, you can assign an IP address to a network interface and assign apartner IP address that the interface takes over if a failover occurs.

Steps

1. Enter the IP address for the interface you are configuring.

For example:

Please enter the IP address for Network Interface e0 []:nnn.nn.nn.nnn

:nnn.nn.nn.nnn is the local address for the node you are configuring.

Note: The addresses for the local node and partner node can reside on different subnetworks.

2. Enter the netmask for the interface you are configuring, or press Return if the default value iscorrect.

For example:

Please enter the netmask for Network Interface e1 [255.255.0.0]:

3. Specify that this interface is to take over a partner IP address.

For example:

Should interface e1 take over a partner IP address during failover? [n]: y

4. Enter the IP address or interface name of the partner.

For example:

Please enter the IP address or interface name to be taken over by e1 []: :nnn.nn.nn.nnn

Note: If the partner is a VIF, you must use the interface name.

Note: The addresses for the local node and partner node can reside on different subnetworks.


Configuring dedicated interfaces with setupYou can assign a dedicated IP address to a network interface, so that the interface does not have apartner IP address.

About this task

This procedure is performed during setup of the storage system.

Steps

1. Enter the IP address for the interface you are configuring.

For example:

Please enter the IP address for Network Interface e0 []::nnn.nn.nn.nnn

:nnn.nn.nn.nnn is the local address for the node you are configuring.

2. Enter the netmask for the interface you are configuring, or press Enter if the default value iscorrect.

For example:


3. Specify that this interface does not take over a partner IP address.

For example:

Should interface e1 take over a partner IP address during failover? [n]: n

Configuring standby interfaces with setupYou can assign a standby IP address to a network interface, so that the interface does not have apartner IP address.

About this task

This procedure is performed during setup of the storage system.

Steps

1. Do not enter an IP address for a standby interface; press Return.

For example:

Please enter the IP address for Network Interface e0 []:

2. Enter the netmask for the interface you are configuring, or press Return if the default value iscorrect.

For example:



3. Specify that this interface is to take over a partner IP address.

For example:

Should interface e1 take over a partner IP address during failover? [n]: y

Enabling licensesYou must enable the required licenses for your type of active/active configuration.

Before you begin

The licenses you need to add depend on the type of your active/active configuration. The followingtable outlines the required licenses for each configuration.

Note: If your system is a V-Series system, you must enable the v-series license on each node in theactive/active configuration.

Configuration type Required licenses

Standard active/active configuration cluster

Mirrored active/active configuration • cluster• syncmirror_local

MetroCluster • cluster• syncmirror_local• cluster_remote

Steps

1. Enter the following command on both node consoles for each required license:

license add license-code

license-code is the license code you received for the feature.

2. Enter the following command to reboot both nodes:

reboot

3. Enter the following command on the local node console:

cf enable

4. Verify that controller failover is enabled by entering the following command on each nodeconsole:


cf status

Cluster enabled, filer2 is up.

Setting options and parametersOptions help you maintain various functions of your node, such as security, file access, and networkcommunication. During takeover, the value of an option might be changed by the node doing thetakeover. This can cause unexpected behavior during a takeover. To avoid unexpected behavior,specific option values must be the same on both the local and partner node.

Option types for active/active configurationsSome options must be the same on both nodes in the active/active configuration, while some can bedifferent, and some are affected by failover events.

In an active/active configuration, options are one of the following types:

• Options that must be the same on both nodes for the active/active configuration to functioncorrectly

• Options that might be overwritten on the node that is failing overThese options must be the same on both nodes to avoid losing system state after a failover.

• Options that should be the same on both nodes so that system behavior does not change duringfailover

• Options that can be different on each node

Note: You can find out whether an option must be the same on both nodes of an active/activeconfiguration from the comments that accompany the option value when you enter the optioncommand. If there are no comments, the option can be different on each node.

Setting matching node optionsBecause some Data ONTAP options need to be the same on both the local and partner node, youneed to check these options with the options command on each node and change them as necessary.

Steps

1. View and note the values of the options on the local and partner nodes, using the followingcommand on each console:

options

The current option settings for the node are displayed on the console. Output similar to thefollowing is displayed:autosupport.doit DONTautosupport.enable on

2. Verify that the options with comments in parentheses are set to the same value for both nodes.The comments are as follows:


Value might be overwritten in takeoverSame value required in local+partnerSame value in local+partner recommended

3. Correct any mismatched options using the following command:options option_name option_value

Note: See the na_options man page for more information about the options.

Parameters that must be the same on each nodeLists the parameters that must be the same so that takeover is smooth and data is transferred betweenthe nodes correctly.

The parameters listed in the following table must be the same so that takeover is smooth and data istransferred between the nodes correctly.

Parameter... Setting for...

date date, rdate

NDMP (on or off) ndmp (on or off)

route table published route

route enabled routed (on or off)

Time zone timezone

Disabling the change_fsid option in MetroCluster configurationsIn a MetroCluster configuration, you can take advantage of the change_fsid option in DataONTAP to simplify site takeover when the cf forcetakeover -d command is used.

About this task

In a MetroCluster configuration, if a site takeover initiated by the cf forcetakeover -dcommand occurs, the following happens:

• Data ONTAP changes the file system IDs (FSIDs) of volumes and aggregates because ownershipchanges.

• Because of the FSID change, clients must remount their volumes if a takeover occurs.• If using Logical Units (LUNs), the LUNs must also be brought back online after the takeover.

To avoid the FSID change in the case of a site takeover, you can set the change_fsid option to off(the default is on). Setting this option to off has the following results if a site takeover is initiated bythe cf forcetakeover -d command:

• Data ONTAP refrains from changing the FSIDs of volumes and aggregates.• Users can continue to access their volumes after site takeover without remounting.• LUNs remain online.


Caution: If the option is set to off, any data written to the failed node that did not get written tothe surviving node's NVRAM is lost. Disable the change_fsid option with great care.

Step

1. Enter the following command to disable the change_fsid option:

options cf.takeover.change_fsid off

By default, the change_fsid option is enabled (set to on ).

Related concepts


Clarification of when data loss can occur when the change_fsid option is enabled

Ensure that you have a good understanding of when data loss can occur before you disable thechange_fsid option. Disabling this option can create a seamless takeover for clients in the event ofa disaster, but there is potential for data loss.

If both the ISLs between the two sites in a fabric MetroCluster go down, then both the systemsremain operational. However, in that scenario, client data is written only to the local plex and theplexes become unsynchronized.

If, subsequently, a disaster occurs at one site, and the cf forcetakeover -d command is issued,the remote plex which survived the disaster is not current. With the change_fsid option set to off,clients switch to the stale remote plex without interruption.

If the change_fsid option is set to on, the system changes the fsids when the cf forcetakeover-d is issued, so clients are forced to remount their volumes and can then check for the integrity of thedata before proceeding.

Verifying and setting the HA state of a32xx controller module and chassis32xx controller modules recognize that they are in an active/active configuration based on HA stateinformation in the controller module and chassis PROMs. If that state information is incorrect(possibly after a chassis or controller module replacement), you can verify the state, and, ifnecessary, update the state.

About this task

• The ha-config show and ha-config modify commands are Maintenance mode commands.• The ha-config command only applies to the local controller module and, in the case of a dual-

chassis active/active configuration, the local chassis.To ensure consistent HA state information throughout the configuration, you must also run thesecommands on the partner controller module and chassis, if necessary.


Steps

1. Boot into Maintenance mode.

2. Enter the following command to display the HA state of the local controller module and chassis:

ha-config show

3. If necessary, enter the following command to set the HA state of the controller module:

ha-config modify controller ha_state

ha_state is ha or non-ha.

The HA state of the controller module is changed.

4. If necessary, enter the following command to set the HA state of the chassis:

ha-config modify chassis ha_state

ha_state is ha or non-ha .

The HA state of the chassis is changed.

5. Repeat the preceding steps on the partner controller module and chassis, if necessary.

Configuring hardware-assisted takeoverYou can use the hardware assisted takeover option to speed up takeover times. The option uses theremote management card to quickly communicate local status changes to the partner node, and hasconfigurable parameters.

Hardware-assisted takeover

Hardware-assisted takeover enables systems with remote management cards to improve the speedwith which takeover events are detected, thereby speeding up the takeover time.

When enabled, hardware-assisted takeover takes advantage of the remote management cardcapabilities to detect failures on the local machine that could require a takeover. If a failure isdetected, the card sends an alert to the partner node and, depending on the type of failure, the partnerperforms the takeover. These alerts can speed takeover because the Data ONTAP takeover processon the partner does not have to take the time to verify that the failing system is no longer giving aheartbeat and confirm that a takeover is actually required.

The hardware-assisted takeover option (cf.hw_assist) is enabled by default.

Requirements for hardware-assisted takeover

The hardware-assisted takeover feature is available only on systems that support Remote LANModules (RLMs) and have the RLMs installed and set up. The remote management card providesremote platform management capabilities, including remote access, monitoring, troubleshooting,logging, and alerting features.

Although a system with an RLM on both nodes provides hardware-assisted takeover on both nodes,hardware-assisted takeover is also supported on active/active configurations in which only one of the


two systems has an installed RLM. The RLM does not have to be installed on both nodes in theactive/active configuration. The RLM can detect failures on the system in which it is installed andprovide faster takeover times if a failure occurs on the system with the RLM.

See the Data ONTAP System Administration Guide for information about setting up the RLM.

System events detected by remote management

A number of events can be detected by the remote management card and generate an alert.Depending on the type of alert received, the partner node initiates takeover.

Alert Takeover initiatedupon receipt?

Description

power_loss Yes Power loss on the node. The remote management cardhas a power supply that maintains power for a shortperiod after a power loss, allowing it to report thepower loss to the partner.

l2_watchdog_reset Yes L2 reset detected by the system watchdog hardware.

power_off_via_rlm Yes The remote management card was used to power offthe system.

power_cycle_via_rlm

Yes The remote management card was used to cycle thesystem power off and on.

reset_via_rlm Yes The remote management card was used to reset thesystem.

abnormal_reboot No Abnormal reboot of the node.

loss_of_heartbeat No Heartbeat message from the node no longer receivedby the remote management card.

Note: This does not refer to the heartbeat messagesbetween the nodes in the active/active configurationbut the heartbeat between the node and its localremote management card.

periodic_message No Periodic message sent during normal hardware-assistedtakeover operation.

test No Test message sent to verify hardware-assisted takeoveroperation.


Disabling and enabling the hardware-assisted takeover option

Hardware-assisted takeover is enabled by default on systems that use an RLM. Hardware-assistedtakeover speeds the takeover process by using the RLM to quickly detect potential takeover eventsand alerting the partner node.

Step

1. Enter the following command to disable or enable the cf.hw_assist option:

options cf.hw_assist.enable off

options cf.hw_assist.enable on

Setting the partner address for hardware-assisted takeover

The cf.hw_assist.partner.address option enables you to change the partner address used bythe hardware-assisted takeover process on the remote management card. The default is the IP addresson the e0a port of the partner. On a31xx system, if the partner's e0M interface has been configured,the IP address of the e0M interface is used. If the e0M interface has not been configured, e0a is used.

Step

1. Enter the following command to set the IP address or host name to which the hardware failurenotification is sent:

options cf.hw_assist.partner.address address_or_hostname

Note: The hardware assisted takeover feature does not support IPv6 addresses when specifyingthe partner IP address in the cf.hw_assist.partner.address address_or_hostnameoption.

If a host name is specified, the host name is resolved when this command is issued.

Setting the partner port for hardware-assisted takeover

When hardware-assisted takeover is enabled, the RLM sends hardware failure notifications to thepartner. The cf.hw_assist.partner.port option enables you to change the partner port. Thedefault is 4444.

Step

1. Enter the following command to set the partner port to which the hardware failure notification issent:

options cf.hw_assist.partner.port port_number


Configuration of network interfacesIf you didn't configure interfaces during system setup, you need to configure them manually toensure continued connectivity during failover.

What the networking interfaces doWhen a node in an active/active configuration fails, the surviving node must be able to assume theidentity of the failed node on the network. Networking interfaces allow individual nodes in theactive/active configuration to maintain communication with the network if the partner fails.

See the Data ONTAP Network Management Guide for a description of available options and thefunction each performs.

Note: You should always use multiple NICs with VIFs to improve networking availability for bothstand-alone storage systems and systems in an active/active configuration.

IPv6 considerations in an active/active configurationWhen enabled, IPv6 provides features such as address autoconfiguration. Using these IPv6 featuresrequires an understanding of how these features work with the active/active configurationfunctionality.

For additional information about IPv6, see the Data ONTAP Administration Guide.

Configuration requirements for using IPv6

To use IPv6 in an active/active configuration, IPv6 must be enabled on both nodes. If a node thatdoes not have IPv6 enabled attempts to take over a node using IPv6, the IPv6 addresses configuredon the partner’s interfaces are lost because the takeover node does not recognize them.

Using the ifconfig command

When using the ifconfig command with IPv4, the partner's interface can be mapped to a localinterface or the partner's IP address. When using IPv6, you must specify the partner interface, not anIP address.

Generation of addresses during takeover

For manually configured IPv6 addresses, during takeover, the mechanism used to configure partner'sIP address remains same as in the case of IPv4.

For link-local auto-configured IPv6 addresses, during takeover, the address is auto-generated basedon the partner's MAC address.

Prefix-based auto-configured addresses are also generated during takeover, based on the prefixes inrouter advertisements (RAs) received on the local link and on the partner's MAC address.


Duplicate Address Detection (DAD) is performed on all IPv6 partner addresses during takeover. Thiscan potentially keep the addresses in tentative state for some amount of time.

IPv6 and hardware-assisted takeover

The hardware assisted takeover feature does not support IPv6 addresses when specifying the partnerIP address in the cf.hw_assist.partner.address address_or_hostname option.

Configuring network interfaces for active/active configurationsConfiguring network interfaces requires that you understand the available configurations for takeoverand that you configure different types of interfaces (shared, dedicated, and standby) depending onyour needs.

Understanding interfaces in an active/active configuration

You can configure three types of interfaces on nodes in an active/active configuration.

Shared, dedicated, and standby interfaces

These different types of interfaces have different roles in normal and takeover mode.

The following table lists the three types of interface configurations that you can enable in an active/active configuration.

Interface type Description

Shared This type of interface supports both the local andpartner nodes. It contains both the local nodeand partner node IP addresses. During takeover,it supports the identity of both nodes.

Dedicated This type of interface only supports the node inwhich it is installed. It contains the local node IPaddress only and does not participate in networkcommunication beyond local node supportduring takeover. It is paired with a standbyinterface.

Standby This type of interface is on the local node, butonly contains the IP address of the partner node.It is paired with a dedicated interface.

Note: Most active/active configuration interfaces are configured as shared interfaces because theydo not require an extra NIC.


Interface roles in normal and takeover modes

You can configure shared, dedicated, and standby interfaces in an active/active configuration. Eachtype has a different role in normal and takeover mode.

The following table shows the role of each interface type in normal and takeover mode.

Interface type Normal mode Takeover mode

Shared Supports the identity of thelocal node

Supports the identity of boththe local node and the failednode

Dedicated Supports the identity of thelocal node

Supports the identity of thelocal node

Standby Idle Supports the identity of thefailed node

Takeover configuration with shared interfaces

You can configure two NICs on to provide two shared interfaces to each node.

In the following configuration illustration, you use two NICs to provide the two interfaces.

If Node 1 fails, interface e0 on Node 1 stops functioning, but the secondary address on e0 on Node 2handles the Node 1 network connection with the 230 network.

If Node 2 fails, e0 on Node 2 stops functioning, but e0 on Node 1 substitutes for the failed interfaceand handles the Node 2 network connection with the 230 network.


Takeover configuration with dedicated and standby interfaces

With two NICs on each node, one can provide a dedicated interface and the other can act as astandby interface.

In the following configuration illustration, you use two NICs for each interface, one on each storagesystem. One NIC acts as a dedicated interface and the other acts as a standby interface.

If Node 1 fails, interface e0 on Node 1 stops functioning, but e0 on Node 2 substitutes for the failedinterface and handles the Node 1 network connection with the 230 network.

If Node 2 fails, e1 on Node 2 stops functioning, but e1 on Node 1 substitutes for the failed interfaceand handles the Node 2 network connection with the 230 network.

Interface types and configurations

This table lists the configurations supported by each type of interface in an active/activeconfiguration.

Interface Shared Dedicated Standby Partnerparameter

Ethernet X X X IP address orinterface name

Gigabit Ethernet X X X IP address orinterface name

Virtual interface X X X Virtual interfacename


Interface Shared Dedicated Standby Partnerparameter

VLAN interface X X X IP address orinterface name

Note: Some storage systems, such as the 31xx systems, include an e0M interface that is dedicatedto management traffic. This port can be partnered in an active/active configuration in the sameway as a regular Ethernet interface.

Making nondisruptive changes to the VIFs

You can use the cf takeover and cf giveback commands to make changes to VIFs in the active/active configuration in a nondisruptive manner.

About this task

Changes to the /etc/rc file require a reboot to make the changes effective. You can use the cftakeover and cf giveback commands to take over one node in the active/active configuration,causing it to reboot while its storage is taken over by the partner.

Steps

1. Edit the /etc/rc file on the desired node to modify the VIFs.

See the Data ONTAP Network Management Guide for more information about configuring VIFs.

2. From the partner node (the partner of the node on which you performed step 1), enter thefollowing command:

cf takeover

3. Enter the following command:

cf giveback

The node on which the changes were made reboots and its /etc/rc file is reread. The rc file isresponsible for creating the VIFs.

4. Repeat these steps, making any required changes to the /etc/rc file on the partner node.

Configuring dedicated and standby interfaces

You can configure dedicated and standby interfaces for an active/active configuration, two on eachnode, so that even in the event of a takeover each node still has a dedicated interface.

Before you begin

Both nodes in the active/active configuration must have interfaces that access the same collection ofnetworks and subnetworks.


You must gather the following information before configuring the interfaces:

• The IP address for both the local node and partner node.

Note: For MetroCluster configurations, if you use the /etc/mcrc file and enable thecf.takeover.use.mcrc_file, the addresses for the local node and partner node can resideon different subnetworks.

• The netmask for both the local node and partner node.• The MTU size for both the local node and partner node. The MTU size must be the same on both

the local and partner interface.

Note: You should always use multiple NICs with VIFs to improve networking availability for bothstand-alone storage systems and systems in an active/active configuration.

About this task

Keep in mind that you can use interface names for specifying all interfaces.

If you configured your interfaces using setup when you first applied power to your storage systems,you do not need to configure them again.

Note: For information about configuring an active/active configuration to use FC, see the DataONTAP Block Access Management Guide for iSCSI and FC.

Steps

1. On nodeA, enter the following command on the command line and also enter it in the /etc/rcfile, so that the command is permanent:

ifconfig interfaceA1addressA1 {other_options}

interfaceA1 is the name of the dedicated local interface for nodeA.

addressA1 is the IP address of the dedicated local interface for nodeA.

other_options denotes whatever other options are needed to correctly configure the interfacein your network environment.

The dedicated local interface for nodeA is configured.

2. Also on nodeA, enter the following command on the command line and in the /etc/rc file:

ifconfig interfaceA2 partner addressB1

interfaceA2 is the name of the standby interface for nodeA.

addressB1

Note: When you configure virtual interfaces for takeover, you must specify the interface nameand not the IP address.


The standby interface for nodeA is configured to take over the dedicated interface of nodeB ontakeover.

3. On nodeB, enter the following command on the command line and in the /etc/rc file:

ifconfig interfaceB1addressB1 {other_options}

interfaceB1 is the name of the dedicated local interface for nodeB.

addressB1 is the IP address of the dedicated local interface for nodeB.

other_options denotes whatever other options are needed to correctly configure the interfacein your network environment.

The dedicated local interface for nodeB is configured.

4. Also on nodeB, enter the following command on the command line and in the /etc/rc file:

ifconfig interfaceB2 partner addressA1

interfaceB2 is the name of the standby interface on nodeB.

addressA1 is the IP address or interface name of the dedicated interface for nodeA.

Note: When you configure virtual interfaces for takeover, you must specify the interface nameand not the IP address.

The standby interface on nodeB is configured to take over the dedicated interface of nodeA ontakeover.

After you finish

If desired, configure your interfaces for automatic takeover in case of NIC failure.

Configuring partner addresses on different subnets (MetroClusters only)On MetroCluster configurations, you can configure partner addresses on different subnets. To dothis, you must create a separate /etc/mcrc file and enable the cf.takeover.use_mcrc_fileoption. When taking over its partner, the node uses the partner's /etc/mcrc file to configure partneraddresses locally. These addresses will reside on the local subnetwork.

The /etc/mcrc file

The /etc/mcrc file, in conjunction with the cf.takeover.use_mcrc_file option, should beused on MetroCluster configurations in which the partner nodes reside on separate subnetworks.

Normally, when a node (for example, nodeA) takes over its partner (nodeB), nodeA runs nodeB's /etc/rc file to configure interfaces on nodeA to handle incoming traffic for the taken-over partner,nodeB. This requires that the local and partner addresses are on the same subnetwork.

When the cf.takeover.use_mcrc_file option is enabled on nodeA, nodeA will use nodeB's /etc/mcrc file upon takeover, instead of nodeB's /etc/rc file. The ifconfig commands in the /etc/mcrc file can configure IP addresses on nodeA's subnetwork. With the correct ifconfig, virtual


IP (VIP), and routing commands in the /etc/mcrc file, the resulting configuration allows hostsconnecting to nodeB to connect to node A.

Note: The /etc/mcrc file must be maintained manually and kept in sync with the /etc/rc file.

Example /etc/rc and /etc/mcrc files

NodeA's /etc/rc file, which configures its local addresses and a partner address (whichmatches the address configured in NodeB's /etc/mcrc file):

hostname nodeAifconfig e0a 10.1.1.1 netmask 255.255.255.0ifconfig e0a partner 10.1.1.100ifconfig vip add 5.5.5.5route add default 10.1.1.50 1routed onoptions dns.domainname mycompany.comoptions dns.enable onoptions nis.enable offsavecore

NodeA's /etc/mcrc file, which configures a partner address on NodeB's subnetwork:

hostname nodeAifconfig e0a 20.1.1.200 netmask 255.255.255.0ifconfig vip add 5.5.5.5route add default 20.1.1.50 1routed onoptions dns.domainname mycompany.comoptions dns.enable onoptions nis.enable offsavecore

NodeB's /etc/rc file, which configures its local addresses and a partner address (whichmatches the address configured in NodeA's /etc/mcrc file)::

hostname nodeBifconfig e0a 20.1.1.1 netmask 255.255.255.0ifconfig e0a partner 20.1.1.200ifconfig vip add 7.7.7.7route add default 20.1.1.50 1routed onoptions dns.domainname mycompany.comoptions dns.enable onoptions nis.enable offsavecore

NodeB's /etc/mcrc file, which configures a partner address on NodeA's subnetwork:

hostname nodeBifconfig e0a 10.1.1.100 netmask 255.255.255.0


ifconfig vip add 7.7.7.7route add default 10.1.1.50 1routed onoptions dns.domainname mycompany.comoptions dns.enable onoptions nis.enable offsavecore

Creating an /etc/mcrc file

You should create an /etc/mcrc file on each node of your MetroCluster configuration if the nodesare on separate subnetworks.

Steps

1. Create an /etc/mcrc file on one node (nodeA) and place it in the /etc directory.

You might want to create the /etc/mcrc file by copying the /etc/rc file.

Note: The /etc/mcrc file must be configured manually. It is not updated automatically. Itmust include all commands necessary to implement the network configuration on the partnernode in the event the node is taken over by the partner.

2. Enter the following commands in nodeA's /etc/mcrc file:

hostname nodeAifconfig interface MetroCluster-partner-address netmask netmaskifconfig vip add virtual-IP-addressroute add default route-for-MetroCluster-partner-address 1routed onother-required-options

interface is the interface on which the corresponding MetroCluster-partner-addresswill reside.

MetroCluster-partner-address is partner address of nodeB. It corresponds to the partneraddress configured by an ifconfig command in nodeB's /etc/rc file.

virtual-IP-address is the virtual address of the partner (nodeB).

other-required-options denotes whatever other options are needed to correctly configurethe interface in your network environment.

ExampleExample of nodeA's /etc/mcrc file:

hostname nodeAifconfig e0a 20.1.1.200 netmask 255.255.255.0ifconfig vip add 5.5.5.5route add default 20.1.1.50 1routed on


options dns.domainname mycompany.comoptions dns.enable onoptions nis.enable offsavecore

3. Create an /etc/mcrc file on the other node (nodeB) and place it in the /etc directory.

The /etc/mcrc file must include an ifconfig command that configures the address thatcorresponds to the address specified in the partner parameter in the partner node's /etc/rc.

You might want to create the /etc/mcrc file by copying the /etc/rc file.

Note: The /etc/mcrc file must be configured manually. It is not updated automatically. Itmust include all commands necessary to configure the

Enter the result of your step here (optional).

4. Enter the following commands in nodeB's /etc/mcrc file:

hostname nodeBifconfig interface MetroCluster-partner-address netmask netmaskifconfig vip add virtual-IP-addressroute add default route-for-MetroCluster-partner-address 1routed onother-required-options

interface is the interface on which the corresponding MetroCluster-partner-addresswill reside.

MetroCluster-partner-address is partner address of nodeA. It corresponds to the partneraddress configured by an ifconfig command in nodeA's /etc/rc file.

virtual-IP-address is the virtual address of the partner (nodeA).

other-required-options denotes whatever other options are needed to correctly configurethe interface in your network environment.

ExampleExample of nodeB's /etc/mcrc file:

hostname nodeBifconfig e0a 10.1.1.100 netmask 255.255.255.0ifconfig vip add 7.7.7.7route add default 10.1.1.50 1routed onoptions dns.domainname mycompany.comoptions dns.enable onoptions nis.enable offsavecore

Setting the system to use the partner's /etc/mcrc file at takeover

You must enable the cf.takeover.use_mcrc_file option to cause the system to use thepartner's /etc/mcrc in the event that the local system takes over the partner. This allows the partner


IP addresses to reside on separate subnetworks. This option should be set on both nodes in theMetroCluster.

Step

1. Enter the following command on both nodes:

options cf.takeover.use_mcrc_file on

The default is off.

Downloading and running the HA Configuration Checkerutility

You can go on the NetApp Support Site and download the Configuration Checker to check forcommon configuration errors.

Before you begin

You must have rsh access to both nodes.

About this task

You can run the utility, cf-config-check.cgi, as a command from a UNIX shell, or you can install theCommon Gateway Interface (CGI) script on a UNIX Web server and invoke it from a Web browser.

Steps

1. To download and run the HA Configuration Checker, log in to the NetApp Support Site and go toSoftware Downloads > Tools & Utilities. Click “HA Configuration Checker (cf-config-check.cgi).”

2. Follow the directions on the web page for downloading and running the utility.

Testing takeover and givebackAfter you configure all aspects of your active/active configuration, verify that it operates as expected.

Steps

1. Check the cabling on the cluster interconnect cables to make sure that they are secure.

2. Verify that you can create and retrieve files on both nodes for each licensed protocol.

3. Enter the following command from the local node console:

cf takeover


The local node takes over the partner node and gives the following output:takeover completed

4. Test communication between the local node and partner node.

ExampleYou can use the fcstat device_map command to ensure that one node can access the othernode’s disks.

5. Give back the partner node by entering the following command:

cf giveback

The local node releases the partner node, which reboots and resumes normal operation. Thefollowing message is displayed on the console when the process is complete:giveback completed

6. Proceed depending on whether you got the message that giveback was completed successfully.

If takeover and giveback... Then...

Is completed successfully Repeat Step 2 through Step 5 on the partner node

Fails Attempt to correct the takeover or giveback failure


Managing takeover and giveback

An active/active configuration allows one partner to take over the storage of the other, and return thestorage using the giveback operation. Management of the nodes in the active/active configurationdiffers depending on whether one partner has taken over the other, and the takeover and givebackoperations themselves have different options.

This information applies to all active/active configurations regardless of disk shelf type.

How takeover and giveback workTakeover is the process in which a node takes over the storage of its partner. Giveback is the processin which the storage is returned to the partner. You can initiate the processes in different ways. Anumber of things that affect the active/active configuration occur when takeover and giveback takeplace.

When takeovers occurThe conditions under which takeovers occur depend on how you configure the active/activeconfiguration.

Takeovers can be initiated when one of the following conditions occur:

• A node is in an active/active configuration that is configured for immediate takeover on panic,and it undergoes a software or system failure that leads to a panic.

• A node that is in an active/active configuration undergoes a system failure (for example, a loss ofpower) and cannot reboot.

Note: If the storage for a node also loses power at the same time, a standard takeover is notpossible. For MetroClusters, you can initiate a forced takeover in this situation.

• There is a mismatch between the disks, array LUNs, or both that one node can see and those thatthe other node can see.

• One or more network interfaces that are configured to support failover become unavailable.• A node cannot send heartbeat messages to its partner. This could happen if the node experienced

a hardware or software failure that did not result in a panic but still prevented it from functioningcorrectly.

• You halt one of the nodes without using the -f flag.• You initiate a takeover manually.

Managing takeover and giveback | 121

What happens during takeoverWhen a takeover occurs, the unimpaired partner node takes over the functions and disk drives of thefailed node by creating an emulated storage system.

The emulated system performs the following tasks:

• Assumes the identity of the failed node• Accesses the failed node’s disks, array LUNs, or both and serves its data to clients

The partner node maintains its own identity and its own primary functions, but also handles theadded functionality of the failed node through the emulated node.

Note: When a takeover occurs, existing CIFS sessions are terminated. A graceful shutdown of theCIFS sessions is not possible, and some data loss could occur for CIFS users.

What happens after takeoverAfter a takeover occurs, you view the surviving partner as having two identities, its own and itspartner’s, that exist simultaneously on the same storage system. Each identity can access only theappropriate volumes and networks. You can send commands or log in to either storage system byusing the rsh command, allowing remote scripts that invoke storage system commands through aRemote Shell connection to continue to operate normally.

Access with rsh

Commands sent to the failed node through a Remote Shell connection are serviced by the partnernode, as are rsh command login requests.

Access with telnet

If you log in to a failed node through a Telnet session, you see a message alerting you that yourstorage system failed and to log in to the partner node instead. If you are logged in to the partnernode, you can access the failed node or its resources from the partner node by using the partnercommand.

What happens during givebackAfter the partner node is repaired and is operating normally, you can use the giveback command toreturn operation to the partner.

When the failed node is functioning again, the following events can occur:

• You initiate a giveback command that terminates the emulated node on the partner.• The failed node resumes normal operation, serving its own data.• The active/active configuration resumes normal operation, with each node ready to take over for

its partner if the partner fails.


Management of an active/active configuration in normalmode

You manage an active/active configuration in normal mode by performing a number of managementactions.

Monitoring active/active configuration statusYou can use commands on the local node to determine whether the controller failover feature isenabled and whether the other node in the active/active configuration is up.

Step


cf status

Examplenode1>

cf status Cluster enabled, node2 is up.

Note: Data ONTAP can disable controller failover if a software or hardware problem existsthat prevents a successful takeover. In this case, the message returned from the cf statuscommand describes the reason why failover is disabled.

This verifies the link between the nodes and tells you that both filer1 and filer2 are functioningand available for takeover.

Monitoring the hardware-assisted takeover featureYou can check and test the hardware-assisted takeover configuration using the hw_assistcommand. You can also use the command to review statistics relating to hardware-assisted takeover.

Checking status

You can check the status of the hardware-assisted takeover configuration with the cf hw_assiststatus command. It shows the current status for the local and partner nodes.

Step

1. Enter the following command to display the hardware-assisted takeover status:

cf hw_assist status


Example hardware-assisted takeover status

The following example shows output from the cf hw_assist status command:

Local Node Status - ha1

Active: Monitoring alerts from partner(ha2)

port 4004 IP address 172.27.1.14

Partner Node Status - ha2

Active: Monitoring alerts from partner(ha1)

port 4005 IP address 172.27.1.15

Testing the hardware-assisted takeover configuration

You can test the hardware-assisted takeover configuration with the cf hw_assist test command.

About this task

The cf hw_assist test command sends a test alert to the partner. If the alert is received thepartner sends back an acknowledgment, and a message indicating the successful receipt of the testalert is displayed on the console.

Step

1. Enter the following command to test the hardware-assisted takeover configuration:

cf hw_assist test

After you finish

Depending on the message received from the cf hw_assist test command, you might need toreconfigure options so that the active/active configuration and the remote management card areoperating.


Checking hardware-assisted takeover statistics

You can display statistics about hardware-assisted takeovers with the cf hw_assist statscommand.

Step

1. Enter the following command to display or clear the hardware-assisted takeover statistics,respectively:

cf hw_assist stats

cf hw_assist stats clear

Example hardware-assisted takeover statistics

The following example shows output from the cf hw_assist stats command on a systemthat has received a variety of alerts from the partner:

# cf hw_assist: stats

Known hw_assist alerts received from partner

alert type alert event num of alerts

---------- ----------- -------------

system_down post_error 0

system_down power_loss 0

system_down abnormal_reboot 0

system_down l2_watchdog_reset 0

system_down power_off_via_rlm 0

system_down power_cycle_via_rlm 0

system_down reset_via_rlm 0

keep_alive loss_of_heartbeat 0

keep_alive periodic_message 18

test test 6

Unknown hw_assist alerts received from partner

Partner nvramid mismatch alerts 5

Shared secret mismatch alerts 10

Unknown alerts 23


Number of times hw_assist alerts throttled: 3

Description of active/active configuration status messagesThe cf status command displays information about the status of the active/active configuration.

The following table shows some of the messages that the cf status command can display.

Message Meaning

cluster enabled, partner_name is up. The active/active configuration is operatingnormally.

partner_name_1 has taken over partner_name_2. One node took over the other node.

Interconnect not present. The system does not recognize the existence of acluster interconnect adapter.

Interconnect is down. The cluster interconnect adapter cannot accessthe partner. This might be due to cablingproblems or the partner might be down.

Interconnect is up. The cluster interconnect adapter is active andcan transmit data to the partner.

partner_name_1 has detected a mailbox diskerror, takeover of partner_name_2 disabled.

One node cannot access multiple mailbox disks.Check access to both the local and partner rootvolumes and mirrors, if they exist. Also checkfor disk or FC-AL problems or offline storageadapters.

partner_name_2 may be down and has disabledtakeover by partner_name_1.

One node might be down.

Version mismatch. The partner node has an incompatible version ofData ONTAP.

partner_name_1 is attempting takeover ofpartner_name_2. takeover is in module n of Nmodules.

A takeover is being attempted (includesinformation about how far the takeover hasprogressed).

partner_name_1 has taken over partner_name_2,giveback in progress. giveback is in module n ofN modules.

A giveback is being attempted (includesinformation about how far the giveback hasprogressed).

partner_name_1 has taken over partner_name_2,partner_name_2 is ready for giveback.

The takeover node received information that thefailed node is ready for giveback.


Message Meaning

partner_name_1 has taken over partner_name_2,partner_name_2 is ready for giveback.

Automatic giveback is disabled due to exceedingretry count.

The takeover node received information that thefailed node is ready for giveback, but givebackcannot take place because the number of retriesexceeded the limit.

Displaying the partner's nameYou can display the name of the other node with the cf partner command.

Step


cf partner

Note: If the node does not yet know the name of its partner because the active/activeconfiguration is new, this command returns “partner”.

Displaying disk and array LUN information on an active/activeconfiguration

To find out about the disks, array LUNs, or both on both the local and partner node, you can use thesysconfig and aggr status commands, which display information about both nodes.

About this task

For each node, the sysconfig command output displays disks on both channel A and channel B:

• The information about disks on channel A is the same as for storage systems not in an active/active configuration.

• The information about disks on channel B is for hardware only; the sysconfig commanddisplays information about the adapters supporting the disks. The command does not showwhether a disk on channelB is a file system disk, spare disk, or parity disk.

Step

1. Enter one of the following commands:

sysconfig -r

or

aggr status -r


Enabling and disabling takeoverYou might want to use the cf disable command to disable takeover if you are doing maintenancethat typically causes a takeover. You can reenable takeover with the cf enable command after youfinish maintenance.

Step


cf enable|disable

Use cf enable to enable takeover or cf disable to disable takeover.

Note: You can enable or disable takeover from either node.

Enabling and disabling automatic takeover of a panicked partnerA node can be configured so it takes over immediately when its partner panics. This shortens thetime between the initial failure and when service is fully restored, because the takeover can bequicker than recovery from the panic, but the subsequent giveback causes another brief outage.

Steps

1. Ensure that you enabled controller takeover by entering the following command:

cf enable

2. Enter the following command to enable or disable takeover on panic:

options cf.takeover.on_panic [on|off]

on Enables immediate takeover of a failed partner or off to disable immediate takeover. This isthe default value.

off Disables immediate takeover. If you disable this option, normal takeover procedures apply.The node still takes over if its partner panics, but might take longer to do so.

Note: If you enter this command on one node, the value applies to both nodes.

The setting of this option is persistent across reboots.

Halting a node without takeoverYou can halt the node and prevent its partner from taking over.

About this task

You can halt the node and prevent its partner from taking over. For example, you might need toperform maintenance on both the storage system and its disks and want to avoid an attempt by thepartner node to write to those disks.


Step


halt -f

Configuring automatic takeoverYou can control when automatic takeovers happen by setting the appropriate options.

Reasons for automatic takeoverYou can set options to control whether automatic takeovers occur due to different system errors. Insome cases, automatic takeover occurs by default unless you disable the option, and in some casesautomatic takeover cannot be prevented.

Takeovers can happen for several reasons. Some system errors must cause a takeover; for example,when a system in an active/active configuration loses power, it automatically fails over to the othernode.

However, for some system errors, a takeover is optional, depending on how you set up your active/active configuration. The following table outlines which system errors can cause a takeover to occur,and whether you can configure the active/active configuration for that error.

System error Option used to configure Default value Notes

A node undergoes asystem failure andcannot reboot.

cf.takeover.on_failure

set to OnOn You should leave this

option enabled unlessinstructed otherwise bytechnical support.

A node undergoes asoftware or systemfailure leading to apanic.

cf.takeover.on_panic set toOn

Off, unless:

• FC oriSCSI islicensed

• Thesystem isaFAS270.

There is a mismatchbetween the disks,array LUNs, or boththat one node can seeand those that the othernode can see.

cf.takeover.on_disk_

shelf_miscompare set to OnOff


System error Option used to configure Default value Notes

All the networkinterface cards (NICs)or VIFs enabled fornegotiated failover ona node failed.

cf.takeover.on_network_

interface_failure set toOn,cf.takeover.on_network_

interface_failure.policy

set to all_nics

By default,takeover onnetworkfailure isdisabled.

To enable a networkinterface for negotiatedfailover, you use theifconfig if_name

nfo command. Formore information, seethe Data ONTAPMultiStoreManagement Guide.

One or more of theNICs or VIFs enabledfor negotiated failoverfailed.

Note: If interfacesfail on both nodes inthe active/activeconfiguration,takeover won'toccur.


interface_failure set to On


interface_failure.policy

set to any_nic

By default,takeover onnetworkfailure isdisabled.

To enable a networkinterface or VIF fornegotiated failover,you use the ifconfigif_name nfo

command. For moreinformation, see theData ONTAPMultiStoreManagement Guide.

A node fails within 60seconds of booting up.

cf.takeover.on_short_upt

ime set to OnOn Changing the value of

this option on one nodeautomatically updatesthe option on thepartner node.

A node cannot sendheartbeat messages toits partner.

n/a You cannot preventthis condition fromcausing a takeover.

You halt one of thenodes without usingthe -f flag.

n/a You cannot preventthis condition fromcausing a takeover. Ifyou include the -fflag, the takeover isprevented.

You initiate a takeovermanually using the cftakeover command.

n/a You cannot preventthis condition fromcausing a takeover.

Related concepts

How disk shelf comparison takeover works on page 133


Related tasks

Enabling and disabling automatic takeover of a panicked partner on page 128

Commands for performing a manual takeoverLists and describes the commands you can use when initiating a takeover. You can initiate a takeoveron a node in an active/active configuration to perform maintenance on that node while still servingthe data on its disks to users.

You can initiate a takeover on a node in an active/active configuration to perform maintenance onthat node while still serving the data on its disks, array LUNs, or both to users. The following tablelists and describes the commands you can use when initiating a takeover:

Command Description

cf takeover Initiates a takeover of the partner of the localnode. Takeover is aborted if a core dump is inprogress on the partner (if thecf.takeover.on_panic option is set to off).The takeover starts either after the partner haltssuccessfully or after a timeout.

cf takeover -f Initiates an immediate takeover of the partner ofthe local node regardless of whether the othernode is dumping its core. The partner node is notallowed to halt gracefully.

cf forcetakeover Tells the cluster monitor to ignore someconfiguration problems that would otherwiseprevent a takeover, such as unsynchronizedNVRAM due to a faulty cluster interconnectconnection. It then initiates a takeover of thepartner of the local node.


Command Description

cf forcetakeover -d Initiates a takeover of the local partner even inthe absence of a quorum of partner mailboxdisks or partner mailbox LUNs.

The cf forcetakeover -d command is validonly if the cluster_remote license is enabled.

Attention: Use the -d option only after youverify that the partner is down.

Note: The -d option is used in conjunctionwith RAID mirroring to recover fromdisasters in which one partner is not available.For more information, see the Data ONTAPData Protection Online Backup and RecoveryGuide.

cf takeover -n Initiates a takeover for a nondisruptive upgrade.For more information, see the Data ONTAPUpgrade Guide.

Specifying the time period before takeoverYou can specify how long (in seconds) a partner in an active/active configuration can beunresponsive before the other partner takes over.

About this task

Both partners do not need to have the same value for this option. Thus, you can have one partner thattakes over more quickly than the other.

Note: If your active/active configuration is failing over because one of the nodes is too busy torespond to its partner, increase the value of the cf.takeover.detection.seconds option onthe partner.

Step


options cf.takeover.detection.seconds number_of_seconds

The valid values for number_of_seconds are 10 through 180; the default is 15.

Note: If the specified time is less than 15 seconds, unnecessary takeovers can occur, and a coremight not be generated for some system panics. Use caution when assigning a takeover time ofless than 15 seconds.


How disk shelf comparison takeover worksDescribes the way a node uses disk shelf comparison with its partner node to determine if it isimpaired.

When communication between nodes is first established through the cluster interconnect adapters,the nodes exchange a list of disk shelves that are visible on the A and B loops of each node. If, later,a system sees that the loop B disk shelf count on its partner is greater than its local loop A disk shelfcount, the system concludes that it is impaired and prompts its partner to initiate a takeover.

Note: Disk shelf comparison does not function for Active/active configurations using software-based disk ownership, or fabric-attached MetroClusters.

Configuring VIFs or interfaces for automatic takeoverAfter you configure your interfaces or VIFs to allow takeovers and givebacks to be completedsuccessfully, you can also optionally configure them to trigger automatic takeover if any or all ofthem experience a persistent failure.

Steps

1. For every VIF or interface on which you want to enable automatic takeover, enter the followingcommand:

ifconfig interface_name nfo

2. Update the /etc/rc file with the command that you entered so that your changes persist acrossreboots.

3. The default policy is that takeover only occurs if all the NICs or VIFs on a node that areconfigured for automatic takeover fail. If you want takeover to occur if any NIC or VIFconfigured for automatic takeover fails, enter the following command:

options cf.takeover.on_network_interface_failure.policy any_nic

4. Enter the following command to enable takeover on interface failures:

options cf.takeover.on_network_interface_failure enable

Note: If interfaces fail on both nodes in the active/active configuration, takeover won't occur.

Takeover of vFiler units and the vFiler unit limitThe vFiler limit, set with the vfiler limit command, determines how many vFiler units can existon a system. In an active/active configuration, if the two systems have different vFiler limits, somevFiler units might not be taken over in the event of a takeover.

When performing a takeover, a system can take over only the number of vFiler units that werespecified by that system's vFiler unit limit. For example, if the limit is set to 5, the system can onlytake over five vFiler units from the partner. If the partner that is being taken over had a higher vFilerlimit, some vFiler units will not be successfully taken over.


For more information about setting the vFiler limit, see the Data ONTAP MultiStore ManagementGuide.

Managing an active/active configuration in takeover modeYou manage an active/active configuration in takeover mode by performing a number ofmanagement actions.

Determining why takeover occurredYou can use the cf status command to determine why a takeover occurred.

Step

1. At the takeover prompt, enter the following command:

cf status

Result

This command can display the following information:

• Whether controller failover is enabled or disabled• Whether a takeover is imminent due to a negotiated failover• Whether a takeover occurred, and the reason for the takeover

Statistics in takeover modeExplains differences in system statistics when in takeover mode.

In takeover mode, statistics for some commands differ from the statistics in normal mode in thefollowing ways:

• Each display reflects the sum of operations that take place on the takeover node plus theoperations on the failed node. The display does not differentiate between the operations on thetakeover node and the operations on the failed node.

• The statistics displayed by each of these commands are cumulative.• After giving back the failed partner’s resources, the takeover node does not subtract the statistics

it performed for the failed node in takeover mode.• The giveback does not reset (zero out) the statistics.

To get accurate statistics from a command after a giveback, you can reset the statistics asdescribed in the man page for the command you are using.

Note: You can have different settings on each node for SNMP options, but any statistics gatheredwhile a node was taken over do not distinguish between nodes.


Managing emulated nodesAn emulated node is a software copy of the failed node that is hosted by the takeover node. Youaccess the emulated node in partner mode by using the partner command.

Management exceptions for emulated nodesThe management of disks and array LUNs and some other tasks are different when you are managingan emulated node.

You manage an emulated node as you do any other storage system, including managing disks orLUNs, with the following exceptions, which are described in greater detail later in this section:

• An emulated node can access only its own disks or LUNs.• Some commands are unavailable.• Some displays differ from normal displays.

Accessing the emulated node from the takeover nodeYou access the emulated node from the takeover node in takeover mode with the partnercommand.

About this task

You can issue the partner command in two forms:

• Using the partner command without an argumentThis toggles between partner mode, in which you manage the emulated node, and takeover mode,in which you manage the takeover node.

• Using the partner command with a Data ONTAP command as an argumentThis executes the command on the emulated node in partner mode and then returns to takeovermode.

Accessing the remote node using the partner command without arguments

Describes how to use the partner command to toggle between the partner mode, in whichcommands are executed on the partner node, and takeover mode.

Step

1. From the takeover prompt, enter the following command:

partner


Result

The prompt changes to the partner mode prompt, which has the following form:emulated_node/takeover_node>

Example showing the change to partner mode

The following example shows the change from takeover mode to partner mode and back:

filer1(takeover)> partnerLogin from console: filer2Thu Aug 20 16:44:39 GMT [filer1: rc]: Login from console: filer2filer2/filer1> partnerLogoff from console: filer2filer1(takeover)> Thu Aug 20 16:44:54 GMT [filer1: rc]: Logoff from console: filer2filer1(takeover)>

Accessing the takeover node with the partner command with arguments

Describes how to use the partner command with a Data ONTAP command as an argument.

Step

1. From the takeover prompt, enter the following command:

partner command

command is the command you want to initiate on the emulated node.

Example of issuing the partner command with an argument

filer1(takeover)> partner cf statusfiler2 has been taken over by filer1.filer1(takeover)>

Assessing the emulated node remotelyYou can also access the emulated node remotely using a Remote Shell (rsh) connection. You cannotaccess the emulated node using Secure Shell (ssh) or Telnet.

Accessing the emulated node remotely using Remote Shell

You can access the emulated node remotely using a Remote Shell (rsh) connection. You cannotaccess the emulated node using Secure Shell (ssh) or Telnet.

Step



rsh failed_node command

failed_node is the name of the failed node.

command is the Data ONTAP command you want to run.

Example of an rsh command

In the following example, filer2 is the failed node.

rsh filer2 df

Emulated node command exceptionsAlmost all the commands that are available to a takeover node are available on the emulated node.Some commands, however, are either unavailable or behave differently in emulated mode.

Unavailable commands

The following commands are not available on an emulated node:

• cf disable• cf enable• cf forcegiveback• cf forcetakeover• cf giveback• cf takeover• date• halt• ifconfig partner• ifconfig -partner• ifconfig mtusize• license cluster• rdate• reboot• timezone


Commands with different behaviors

Command Difference

ifconfig [interface] Displays the following:

• Emulated interface mappings based on thefailed node's /etc/rc file rather than thetakeover node interface mappings

Note: MetroCluster nodes use the failednode's /etc/mcrc file if thecf.takeover.use_mcrc_file option isenabled.

• Emulated interface names rather than theinterface names of the takeover node

• Only interfaces that have been configured,rather than all interfaces, configured or not,as displayed on the takeover node

mt Uses the tape devices on the takeover nodebecause the failed node has no access to its tapedevices.

netstat -i Appends a plus sign (+) to shared interfaces. Ashared interface is one that has two IP addressesassigned to it: an IP address for the node inwhich it physically resides and an IP address forits partner node in the active/activeconfiguration.

sysconfig When it displays hardware information, thesysconfig command displays information onlyabout the hardware that is attached to thetakeover node. It does not display informationabout the hardware that is attached only to thefailed node. For example, the disk adapterinformation that the partner sysconfig -rcommand displays is about the disk adapters onthe takeover node.

uptime Displays how long the failed node has beendown and the host name of the takeover node.


Command Difference

aggr status When it displays hardware information, theaggr status command displays information onlyabout the hardware that is attached to thetakeover node. It does not display informationabout the hardware that is attached only to thefailed node. For example, the disk adapterinformation that the partner aggr status -rcommand displays is about the disk adapters onthe takeover node.

Performing dumps and restores for a failed nodeYou can use the emulated node and peripheral devices attached to the takeover node to performdumps and restores for the failed node.

Before you begin

Any dump commands directed to the failed node’s tape drives are executed on the takeover node’stape drives. Therefore, any dump commands that you execute using a scheduler, such as the croncommand, succeed only under the following conditions:

• The device names are the same on both nodes in the active/active configuration.• The dump commands for the takeover node and the emulated node are not scheduled to occur

during the same time period; the takeover node and the emulated node cannot access the tapedrives simultaneously.

About this task

Because the peripheral devices for a failed node are inaccessible, you perform dumps and restores fora failed node by using the emulated node (available using the partner command on the takeovernode), making sure that you use a peripheral device attached to the takeover node.

For more information about performing dumps and restores, see the Data ONTAP Data ProtectionTape Backup and Recovery Guide.

Step

1. Issue the backup or restore command, either in partner mode or as an argument in thepartner command.

ExampleIssuing a restore command in partner mode:


filer1 (takeover)> partnerfiler1/filer2> restore [options [arguments]]filer1 (takeover)> partner

ExampleIssuing a restore command as an argument in the partner command:

filer1 (takeover)> partner restore [options [arguments]]

Giveback operationsGiveback can be implemented and configured in a number of different ways. It can also beconfigured to occur automatically.

Performing a manual givebackYou can perform a normal giveback, a giveback in which you terminate processes on the partnernode, or a forced giveback.

Note: Prior to performing a giveback, you must remove failed drives in the taken-over system.

Removing failed disks prior to attempting giveback

For taken-over systems that use disks, you must remove the failed disk or disks prior to attempting toimplement giveback.

Step

1. Remove the failed disks, as described in the Storage Management Guide.

After you finish

When all failed disks are removed or replaced, proceed with the giveback operation.

Initiating normal giveback

You can return control to a taken-over partner with the cf giveback command.

Before you begin

On a fabric-attached MetroCluster, before you undertake the giveback operation, you must rejoin theaggregates on the surviving node and the partner node to reestablish the MetroCluster configuration.

Step

1. Enter the following command on the command line of the takeover node:

cf giveback


Note: If the giveback fails, there might be a process running that prevents giveback. You canwait and repeat the command, or you can use the initiate giveback using the -f option toterminate the processes that are preventing giveback.

After a giveback, the takeover node’s ability to take over its partner automatically is notreenabled until the partner reboots successfully. If the partner fails to reboot, you can enter thecf takeover command to initiate a takeover of the partner manually.

Troubleshooting if giveback fails

If the cf giveback command fails, you should check for system processes that are currentlyrunning and might prevent giveback, check that the cluster interconnect is operational, and check forany failed disks on systems using disks.

Steps

1. For systems using disks, check for and remove any failed disks, as described in the Data ONTAPStorage Management Guide.

2. Check for the messagecf.giveback.disk.check.fail

on the console. Both nodes should be able to detect the same disks. This message indicates thatthere is a disk mismatch: for some reason, one node is not seeing all the disks attached to theactive/active configuration.

3. Check the cluster interconnect and verify that it is correctly connected and operating.

4. Check whether any of the following processes were taking place on the takeover node at the sametime you attempted the giveback:

• Outstanding CIFS sessions• RAID disk additions• Volume creation (traditional volume or FlexVol volume)• Aggregate creation• Disk ownership assignment• Disks being added to a volume (vol add)• Snapshot copy creation, deletion, or renaming• Quota initialization• Advanced mode repair operations, such as wafliron• Storage system panics• Backup dump and restore operations• SnapMirror transfers (if the partner is a SnapMirror destination)• SnapVault restorations• Disk sanitization operations

If any of these processes are taking place, either cancel the process or wait until it is complete,and then try the giveback operation again.


5. If the cf giveback operation still does not succeed, use the cf giveback -f command toforce giveback.

Related tasks

Forcing giveback on page 142

Forcing giveback

Because the takeover node might detect an error condition on the failed node that typically prevents acomplete giveback, such as data not being flushed from NVRAM to the failed node’s disks, you canforce a giveback, if necessary.

About this task

You can use this procedure to force the takeover node to give back the resources of the failed nodeeven if the takeover node detects an error that typically prevents a complete giveback.

Step

1. On the takeover node, enter the following command:

cf forcegiveback

Attention: Use cf forcegiveback only when you cannot get cf giveback to succeed.When you use this command, you risk losing any data committed to NVRAM but not to disk.

If a cifs terminate command is running, allow it to finish before forcing a giveback.

If giveback is interrupted

If the takeover node experiences a failure or a power outage during the giveback process, thegiveback process stops and the takeover node returns to takeover mode when the failure is repairedor the power is restored.

Configuring givebackYou can configure how giveback occurs, setting different Data ONTAP options to improve the speedand timing of giveback.


Option for shortening giveback time

You can shorten the client service outage during giveback by using thecf.giveback.check.partner option. You should always set this option to on.

Setting giveback delay time for CIFS clients

You can specify the number of minutes to delay an automatic giveback before terminating CIFSclients that have open files.

About this task

This option specifies the number of minutes to delay an automatic giveback before terminating CIFSclients that have open files. During the delay, the system periodically sends notices to the affectedclients. If you specify 0, CIFS clients are terminated immediately.

This option is used only if automatic giveback is On.

Step


options cf.giveback.auto.cifs.terminate.minutes minutes

Valid values for minutes are 0 through 999. The default is 5 minutes.

Option for terminating long-running processes

Describes the cf.giveback.auto.terminate.bigjobs option, which, when on, specifies thatautomatic giveback should immediately terminate long-running operations.

The cf.giveback.auto.terminate.bigjobs option, when on, specifies that automatic givebackshould immediately terminate long-running operations (dump/restore, vol verify, and so on) wheninitiating an automatic giveback. When this option is off, the automatic giveback is deferred until thelong-running operations are complete. This option is used only if automatic giveback is On.

Setting giveback to terminate long-running processes

You can set the automatic giveback process to terminate long-running processes that might preventthe giveback.

Step


options cf.giveback.auto.terminate.bigjobs {on|off}

The on argument enables this option. The off argument disables this option. This option is on bydefault.


Configuring automatic givebackYou can enable automatic giveback by using the cf.giveback.auto.enable option.

About this task

You should use the automatic giveback feature with care:

• Do not enable automatic giveback in MetroCluster configurations. Before the giveback operationis undertaken, you must rejoin the aggregates on the surviving node and the partner node toreestablish the MetroCluster configuration. If automatic giveback is enabled, this crucial stepcannot be performed before the giveback.

• You should leave this option disabled unless your clients are unaffected by failover, or you haveprocesses in place to handle repetitive failovers and givebacks.

Step

1. Enable the following option to enable automatic giveback: cf.giveback.auto.enable on.The on value enables automatic giveback. The off value disables automatic giveback. Thisoption is off by default.

Troubleshooting takeover or giveback failuresIf takeover or giveback fails for an active/active configuration, you need to check the cluster statusand proceed based on messages you receive.

Steps

1. Check communication between the local and partner nodes by entering the following commandand observing the messages:

cf status

2. Review the messages and take the appropriate action:

If the error message indicates... Then...

A cluster adapter error Check the cluster adapter cabling. Make sure that the cabling iscorrect and properly seated at both ends of the cable.

That the NVRAM adapter is in thewrong slot number

Check the NVRAM slot number. Move it to the correct slot ifneeded.

A Channel B cabling error Check the cabling of the Channel B disk shelf loops and reseat andtighten any loose cables.


If the error message indicates... Then...

A networking error Check for network connectivity.

See the Data ONTAP MultiStore Management Guide for moreinformation.

3. If you have not already done so, run the HA Configuration Checker script.

4. Correct any errors or differences displayed in the output.

5. Reboot the active/active configuration and rerun the takeover and giveback tests.

6. If you still do not have takeover enabled, contact technical support.

Related tasks

Downloading and running the HA Configuration Checker utility on page 119


Managing DS14mk2 or DS14mk4 FC disk shelvesin an active/active configuration

You must follow specific procedures to add disk shelves to an active/active configuration or aMetroCluster, or to upgrade or replace disk shelf hardware in an active/active configuration.

If your configuration includes SAS disk shelves, refer to the following documents on the NetAppSupport Site at now.netapp.com/NOW/knowledge/docs/hardware/hardware_index.shtml:

• For SAS disk shelf management, see the Installation and Service Guide for your disk shelf model.• For cabling SAS disk shelves in an active/active configuration, see the Universal SAS and ACP

Cabling Guide.

Note: Fabric-attached MetroCluster configurations do not support SAS disk shelves.

Adding DS14mk2 or DS14mk4 FC disk shelves to amultipath HA loop

To add supported DS14mk2 or DS14mk4 FC disk shelves to an active/active configurationconfigured for multipath HA, you need to add the new disk shelf to the end of a loop, ensuring that itis connected to the previous disk shelf and to the controller.

About this task

This task does not apply to SAS disk shelves.

Steps

1. Confirm that there are two paths to every disk by entering the following command:

storage show disk -p

Note: If there are not two paths listed for every disk, this procedure could result in a dataservice outage. Before proceeding, address any issues so that all paths are redundant. If you donot have redundant paths to every disk, you can use the nondisruptive upgrade method(failover) to add your storage.

2. Install the new disk shelf in your cabinet or equipment rack, as described in the DiskShelf14mk2/mk4 Hardware Service Guide.

3. Determine whether disk shelf counting is enabled by entering the following command:

options cf.takeover.on_disk_shelf_miscompare

4. If the disk shelf counting option is set to On, turn it off by entering the following command:

Managing DS14mk2 or DS14mk4 FC disk shelves in an active/active configuration | 147


options cf.takeover.on_disk_shelf_miscompare off

5. Find the last disk shelf in the loop for which you want to add the new disk shelf.

Note: The Channel A Output port of the last disk shelf in the loop is connected back to one ofthe controllers.

Note: In Step 6 you disconnect the cable from the disk shelf. When you do this the systemdisplays messages about adapter resets and eventually indicates that the loop is down. Thesemessages are normal within the context of this procedure. However, to avoid them, you canoptionally disable the adapter prior to disconnecting the disk shelf.

If you choose to, disable the adapter attached to the Channel A Output port of the last diskshelf by entering the following command:

fcadmin config -d <adapter>

<adapter> identifies the adapter by name. For example: 0a.

6. Disconnect the SFP and cable coming from the Channel A Output port of the last disk shelf.

Note: Leave the other ends of the cable connected to the controller.

7. Using the correct cable for a shelf-to-shelf connection, connect the Channel A Output port of thelast disk shelf to the Channel A Input port of the new disk shelf.

8. Connect the cable and SFP you removed in Step 6 to the Channel A Output port of the new diskshelf.

9. If you disabled the adapter in Step 5, reenable the adapter by entering the following command:

fcadmin config -e <adapter>

10. Repeat Step 6 through Step 9 for Channel B.

Note: The Channel B Output port is connected to the other controller.

11. Confirm that there are two paths to every disk by entering the following command:

storage show disk -p

There should be two paths listed for every disk.

12. If disk shelf counting was Off, reenable it by entering the following command:

options cf.takeover.on_disk_shelf_miscompare on

Related tasks

Determining path status for your active/active configuration on page 151


Upgrading or replacing modules in an active/activeconfiguration

In an active/active configuration with redundant pathing, you can upgrade or replace disk shelfmodules without interrupting access to storage.

About this task

These procedures are for DS14mk2 AT, DS14mk2 FC, or DS14mk4 FC disk shelves.

Note: If your configuration includes DS4243 disk shelves, refer to the DS4243 Installation andService Guide on the NetApp Support Site at support.netapp.com.

About the disk shelf modulesA disk shelf module (LRC, ESH, ESH2, ESH4, or AT-FCX) in a DS14, DS14mk2, DS14mk4 FC orDS14mk2 AT includes a SCSI-3 Enclosure Services Processor that maintains the integrity of theloop when disks are swapped and provides signal retiming for enhanced loop stability. Whenupgrading or replacing a module, you must be sure to cable the modules correctly.

The DS14, DS14mk2, DS14mk4 FC or DS14mk2 AT disk shelves support the LRC, ESH, ESH2,ESH4, or AT-FCX modules.

There are two modules in the middle of the rear of the disk shelf, one for Channel A and one forChannel B.

Note: The Input and Output ports on module B on the DS14/DS14mk2/DS14mk4 FC shelf are thereverse of module A.

Restrictions for changing module typesIf you plan to change the type of any module in your active/active configuration, make sure that youunderstand the restrictions.

• You cannot mix LRC and ESH modules in the same loop. Doing so results in loop failure.• You cannot mix LRC, ESH, ESH2 or ESH4 modules in the same loop with AT-FCX modules.• You cannot mix ESH and ESH4 modules in the same loop.• To replace an ESH with an ESH2, you must do the following tasks:

• Upgrade to Data ONTAP 6.4.4 or later, if necessary.• Replace all ESH modules in the same disk shelf with ESH2 modules.



Best practices for changing module typesIf you plan to change the type of any module in your active/active configuration, make sure that youreview the best practice guidelines.

• Whenever you remove a module from an active/active configuration, you need to know whetherthe path you will disrupt is redundant. If it is, you can remove the module without interfering withthe storage system’s ability to serve data. However, if that module provides the only path to anydisk in your active/active configuration, you must take action to ensure that you do not incursystem downtime.

• When you replace a module, make sure that the replacement module’s termination switch is in thesame position as the module it is replacing.

Note: ESH2 and ESH4 modules are self-terminating; this guideline does not apply to ESH2and ESH4 modules.

• If you replace a module with a different type of module, make sure that you also change thecables, if necessary.For more information about supported cable types, see the hardware documentation for your diskshelf.

• Always wait 30 seconds after inserting any module before reattaching any cables in that loop.• ESH2 and ESH4 modules should not be on the same disk shelf loop.

Related concepts

Understanding redundant pathing in active/active configurations on page 150

Testing the modulesYou should test your disk shelf modules after replacing or upgrading them, to ensure that they areconfigured correctly and operating.

Steps

1. Verify that all disk shelves are functioning properly by entering the following command:

environ shelf

2. Verify that there are no missing disks by entering the following command:

aggr status -r

Local disks displayed on the local node should be displayed as partner disks on the partner node,and vice-versa.

3. Verify that you can create and retrieve files on both nodes for each licensed protocol.

Understanding redundant pathing in active/active configurationsSome active/active configurations have two paths from each controller to each of their disk shelvesor array LUNs; this configuration is called a redundant-path or multipath configuration. Active/active


configurations using Multipath HA are redundant-path configurations. Fabric-attached MetroClustersare also redundant-path configurations.

Determining path status for your active/active configurationYou can determine whether any module in your system provides the only path to any disk or arrayLUN by using the storage show disk -p command at your system console.

About this task

If you want to remove a module from your active/active configuration, you need to know whether thepath you will disrupt is redundant. If it is, you can remove the module without interfering with thestorage system’s ability to serve data. On the other hand, if that module provides the only path to anyof the disks or array LUNs in your active/active configuration, you must take action to ensure thatyou do not incur system downtime.

Step

1. Use the storage show disk -p command at your system console.

This command displays the following information for every disk or array LUN in the active/active configuration:

• Primary port• Secondary port• Disk shelf• Bay

Examples for configurations with and without redundant paths

The following example shows what the storage show disk -p command output mightlook like for a redundant-path active/active configuration consisting of FAS systems:PRIMARY PORT SECONDARY PORT SHELF BAY------- ---- --------- ---- ---------0c.112 A 0b.112 B 7 0 0b.113 B 0c.113 A 7 1 0b.114 B 0c.114 A 7 2 0c.115 A 0b.115 B 7 30c.116 A 0b.116 B 7 4 0c.117 A 0b.117 B 7 5 0b.118 B 0c.118 A 7 6 0b.119 B 0c.119 A 7 7 0b.120 B 0c.120 A 7 8 0c.121 A 0b.121 B 7 9 0c.122 A 0b.122 B 7 100b.123 B 0c.123 A 7 11

Notice that every disk (for example, 0c.112/0b.112) has two ports active: one for A and onefor B. The presence of the redundant path means that you do not need to fail over one systembefore removing modules from the system.


Attention: Make sure that every disk or array LUN has two paths. Even in an active/activeconfiguration configured for redundant paths, a hardware or configuration problem cancause one or more disks to have only one path. If any disk or array LUN in your active/active configuration has only one path, you must treat that loop as if it were in a single-pathactive/active configuration when removing modules.

The following example shows what the storage show disk -p command output mightlook like for an active/active configuration consisting of FAS systems that do not useredundant paths:

filer1> storage show disk -pPRIMARY PORT SECONDARY PORT SHELF BAY------- ---- --------- ---- ---------5b.16 B 1 0 5b.17 B 1 1 5b.18 B 1 2 5b.19 B 1 3 5b.20 B 1 4 5b.21 B 1 5 5b.22 B 1 6 5b.23 B 1 7 5b.24 B 1 8 5b.25 B 1 9 5b.26 B 1 10 5b.27 B 1 11 5b.28 B 1 12 5b.29 B 1 13 5b.32 B 2 0 5b.33 B 2 1 5b.34 B 2 2 5b.35 B 2 3 5b.36 B 2 4 5b.37 B 2 5 5b.38 B 2 6 5b.39 B 2 7 5b.40 B 2 8 5b.41 B 2 9 5b.42 B 2 10 5b.43 B 2 11 5b.44 B 2 12 5b.45 B 2 13

For this active/active configuration, there is only one path to each disk. This means that youcannot remove a module from the configuration, thereby disabling that path, without firstperforming a takeover.


Hot-swapping a moduleYou can hot-swap a faulty disk shelf module, removing the faulty module and replacing it withoutdisrupting data availability.

About this task

When you hot-swap a disk shelf module, you must ensure that you never disable the only path to adisk, which results in a system outage.

Attention: If there is newer firmware in the /etc/shelf_fw directory than that on thereplacement module, the system automatically runs a firmware update. On non-multipath HA AT-FCX installations, multipath HA configurations running versions of Data ONTAP prior to 7.3.1,and non-RoHS modules, this firmware update causes a service interruption.

Steps

1. Verify that your storage system meets the minimum software requirements to support the diskshelf modules that you are hot-swapping.

See the DiskShelf14, DiskShelf14mk2 FC, or DiskShelf14mk2 AT Hardware Guide for moreinformation.

2. Determine which loop contains the module you are removing, and determine whether any disksare single-pathed through that loop.

3. If any disks use this loop for their only path to a controller, complete the following steps:

a. Follow the cables from the module you want to replace back to one of the nodes, called NodeA.

b. At the Node B console, enter the following command:

cf takeover

c. Wait for takeover to be complete and make sure that the partner node, or Node A, reboots andis waiting for giveback.

Any module in the loop that is attached to Node A can now be replaced.

4. Note whether the module you are replacing has a terminate switch. If it does, set the terminateswitch of the new module to the same setting.

Note: The ESH2 and ESH4 are self-terminating and do not have a terminate switch.

5. Put on the antistatic wrist strap and grounding leash.

6. Disconnect the module that you are removing from the Fibre Channel cabling.

7. Using the thumb and index finger of both hands, press the levers on the CAM mechanism on themodule to release it and pull it out of the disk shelf.


8. Slide the replacement module into the slot at the rear of the disk shelf and push the levers of thecam mechanism into place.

Attention: Do not use excessive force when sliding the module into the disk shelf; you mightdamage the connector.

Wait 30 seconds after inserting the module before proceeding to the next step.

9. Recable the disk shelf to its original location.

10. Check the operation of the new module by entering the following command from the console ofthe node that is still running:

environ shelf

The node reports the status of the modified disk shelves.

11. If you performed a takeover previously, complete the following steps:

a. At the console of the takeover node, return control of Node B’s disk shelves by entering thefollowing command:

cf giveback

b. Wait for the giveback to be completed before proceeding to the next step.

12. Test the replacement module.

13. Test the configuration.

Related concepts

Best practices for changing module types on page 150

Related tasks

Determining path status for your active/active configuration on page 151


Disaster recovery using MetroCluster

In situations such as prolonged power outages or natural disasters, you can use the optionalMetroCluster feature of Data ONTAP to provide a quick failover to another site that contains anearly real time copy of the data at the disaster site.


Conditions that constitute a disasterThe disaster recovery procedure is an extreme measure that you should use only if the failure disruptsall communication from one MetroCluster site to the other for a prolonged period of time.

The following are examples of disasters that could cause such a failure:

• Fire• Earthquake• Prolonged power outages at a site• Prolonged loss of connectivity from clients to the storage systems at a site

Ways to determine whether a disaster occurredYou should declare a disaster only after using predefined procedures to verify that service cannot berestored.

It is critical that you follow a predefined procedure to confirm that a disaster occurred. The procedureshould include determining the status of the disaster site by:

• Using external interfaces to the storage system, such as the following:

• ping command to verify network connectivity• Remote shell• FilerView administration tool

• Using network management tools to verify connectivity to the disaster site• Physically inspecting the disaster site, if possible

You should declare a disaster only after verifying that service cannot be restored.

Failures that do not require disaster recoveryIf you can reestablish the MetroCluster connection after fixing the problem, you should not performthe disaster recovery procedure.

Do not perform the disaster recovery procedure for the following failures:

Disaster recovery using MetroCluster | 155

• A failure of the cluster interconnect between the two sites. This can be caused by the following:

• Failure of the interconnect cable• Failure of one of the FC-VI adapters• If using switches, a failure of the SFP connecting a node to the switch

With this type of failure, both nodes remain running. Automatic takeover is disabled becauseData ONTAP cannot synchronize the NVRAM logs. After you fix the problem and reestablishthe connection, the nodes resynchronize their NVRAM logs and the MetroCluster returns tonormal operation.

• The storage from one site (site A) is not accessible to the node at the other site (site B). This canbe caused by the following:

• Failure of any of the cables connecting the storage at one site to the node at the other site orswitch

• If using switches, failure of any of the SFPs connecting the storage to the switch or the nodeto the switch

• Failure of the Fibre Channel adapter on the node• Failure of a storage disk shelf (disk shelf module; power; access to disk shelves; and so on)

With this type of failure, you see a “mailbox disk invalid” message on the console of the storagesystem that cannot see the storage. After you fix the problem and reestablish the connection, theMetroCluster returns to normal operation.

• If you are using switches, the inter-switch link between each pair of switches fails.With this type of failure, both nodes remain running. You see a “mailbox disk invalid” messagebecause a storage system at one site cannot see the storage at the other site. You also see amessage because the two nodes cannot communicate with each other. After you fix the problemand reestablish the connection, the nodes resynchronize their NVRAM logs and the MetroClusterreturns to normal operation.

Recovering from a disasterAfter determining that there is a disaster, you should take steps to recover access to data, fixproblems at the disaster site, and re-create the MetroCluster configuration.

About this task

Complete the following tasks in the order shown.

Attention: If for any reason the primary node has data that was not mirrored to the secondary priorto the execution of the cf forcetakeover –d command, data could be lost. Do notresynchronize the original disks of the primary site for a SnapLock volume until an additionalbackup has been made of those disks to assure availability of all data. This situation could arise,for example, if the link between the sites was down and the primary node had data written to it inthe interim before the cf forcetakeover –d command was issued.


For more information about backing up data in SnapLock volumes using SnapMirror, see the DataONTAP Archive and Compliance Management Guide.

Steps

1. Restricting access to the disaster site node on page 157

2. Forcing a node into takeover mode on page 158

3. Remounting volumes of the failed node on page 158

4. Recovering LUNs of the failed node on page 159

5. Fixing failures caused by the disaster on page 160

6. Reestablishing the MetroCluster configuration on page 161

Restricting access to the disaster site nodeYou must restrict access to the disaster site node to prevent the node from resuming service. If youdo not restrict access, you risk the possibility of data corruption.

About this task

You can restrict access to the disaster site node in the following ways:

• Turning off power to the disaster site node.• Manually fencing off the node.

Steps

1. Restricting access to the node by turning off power on page 157

2. Restricting access to the node by fencing off on page 157

Restricting access to the node by turning off power

This is the preferred method for restricting access to the disaster site node. You can perform this taskat the disaster site or remotely, if you have that capability.

Step

1. Switch off the power at the back of the storage system.

Restricting access to the node by fencing off

You can use manual fencing as an alternative to turning off power to the disaster site node. Themanual fencing method restricts access using software and physical means.

Steps

1. Disconnect the cluster interconnect and Fibre Channel adapter cables of the node at the survivingsite.


2. Use the appropriate fencing method depending on the type of failover you are using:

If you are using... Then fencing is achieved by...

Applicationfailover

Using any application-specified method that either prevents the application fromrestarting at the disaster site or prevents the application clients from accessing theapplication servers at the disaster site. Methods can include turning off theapplication server, removing an application server from the network, or any othermethod that prevents the application server from running applications.

IP failover Using network management procedures to ensure that the storage systems at thedisaster site are isolated from the external public network.

Forcing a node into takeover modeIf a disaster has occurred, you can force the surviving node into takeover mode, so that the survivingnode serves the data of the failed node.

Step

1. Enter the following command on the surviving node:

cf forcetakeover -d

Result

Data ONTAP causes the following to occur:

• The surviving node takes over the functions of the failed node.• The mirrored relationships between the two plexes of mirrored aggregates are broken, thereby

creating two unmirrored aggregates. This is called splitting the mirrored aggregates.

The overall result of using the cf forcetakeover -d command is that a node at the surviving siteis running in takeover mode with all the data in unmirrored aggregates.

Remounting volumes of the failed nodeIf the cf.takeover.change_fsid option is set to on, you must remount the volumes of the failednode because the volumes are accessed through the surviving node.

About this task

For more information about mounting volumes, see the Data ONTAP File Access and ProtocolsManagement Guide.

Note: You can disable the change_fsid option to avoid the necessity of remounting the volumes.


Steps

1. On an NFS client at the surviving site, create a directory to act as a mount point by entering thefollowing command.

mkdir directory_path

Example

mkdir /n/toaster/home

2. Mount the volume by entering the following command.

mount volume_name

Example

mount toaster:/vol/vol0/home /n/toaster/home

Related tasks


Recovering LUNs of the failed nodeYou must actively track whether LUNs are online or offline in a MetroCluster configuration. If thecf.takeover.change_fsid option is set to on, and there is a disaster, all LUNs in the aggregatesthat were mirrored at the surviving site are offline. You can’t determine if they were online prior tothe disaster unless you track their state.

About this task

If you have a MetroCluster configuration, you must actively track the state of LUNs (track whetherthey are online or offline) on the node at each site. If there is a failure to a MetroClusterconfiguration that qualifies as a disaster and the node at one site is inaccessible, all LUNs in theaggregates that were mirrored at the surviving site are offline. There is no way to distinguish theLUNs that were offline before the disaster from the LUNs that were online before the disaster unlessyou have been tracking their status.

When you recover access to the failed node’s LUNs, it is important to bring back online only theLUNs that were online before the disaster. To avoid igroup mapping conflicts, do not bring a LUNonline if it was offline before the disaster. For example, suppose you have two LUNs with IDs of 5mapped to the same igroup, but one of these LUNs was offline before the disaster. If you bring thepreviously offline LUN online first, you cannot bring the second LUN online because you cannothave two LUNs with the same ID mapped to the same host.

Note: You can disable the change_fsid option to avoid the necessity of remounting the volumes.

Steps

1. Identify the LUNs that were online before the disaster occurred.


2. Make sure that the LUNs are mapped to an igroup that contains the hosts attached to thesurviving node.

For more information about mapping LUNs to igroups, see your Data ONTAP Block AccessManagement Guide for iSCSI and FC.

3. On the surviving node, enter the following command:

lun online lun-path ...

lun-path is the path to the LUN you want to bring online. You can specify more than one pathto bring multiple LUNs online.

Example

lun online /vol/vol1/lun5

Example

lun online /vol/vol1/lun3 /vol/vol1/lun4

Note: After you bring LUNs back online, you might have to perform some application or host-side recovery procedures. For example, the File System Identifiers (FSIDs) are rewritten,which can cause the LUN disk signatures to change. For more information, see thedocumentation for your application and for your host operating system.

Fixing failures caused by the disasterYou need to fix the failures caused by the disaster, if possible. For example, if a prolonged poweroutage to one of the MetroCluster sites caused the failure, restoring the power fixes the failure.

Before you begin

You cannot fix failures if the disaster causes a site to be destroyed. For example, a fire or anearthquake could destroy one of the MetroCluster sites. In this case, you fix the failure by creating anew MetroCluster-configured partner at a different site.

Step

1. Fix the failures at the disaster site.

After you finish

After the node at the surviving site can see the disk shelves at the disaster site, Data ONTAP renamesthe mirrored aggregates that were split, and the volumes they contain, by adding a number inparenthesis to the name. For example, if a volume name was vol1 before the disaster and the split, therenamed volume name could be vol1(1).


Reestablishing the MetroCluster configurationDescribes how to reestablish a MetroCluster after a disaster, depending on the state of the mirroredaggregate at the time of the takeover.

About this task

Depending on the state of a mirrored aggregate before you forced the surviving node to take over itspartner, you use one of two procedures to reestablish the MetroCluster configuration:

• If the mirrored aggregate was in a normal state before the forced takeover, you can rejoin the twoaggregates to reestablish the MetroCluster configuration. This is the most typical case.

• If the mirrored aggregate was in an initial resynchronization state (level-0) before the forcedtakeover, you cannot rejoin the two aggregates. You must re-create the synchronous mirror toreestablish the MetroCluster configuration.

Rejoining the mirrored aggregates to reestablish a MetroCluster

Describes how to rejoin the mirrored aggregates if the mirrored aggregate was in a normal statebefore the forced takeover.

About this task

Attention: If you attempt a giveback operation prior to rejoining the aggregates, you might causethe node to boot with a previously failed plex, resulting in a data service outage.

Steps

1. Validate that you can access the remote storage by entering the following command:

aggr status -r

2. Turn on power to the node at the disaster site.

After the node at the disaster site boots, it displays the following message:Waiting for Giveback...

3. Determine which aggregates are at the surviving site and which aggregates are at the disaster siteby entering the following command:

aggr status

Aggregates at the disaster site show plexes that are in a failed state with an out-of-date status.Aggregates at the surviving site show plexes as online.

4. If aggregates at the disaster site are online, take them offline by entering the following commandfor each online aggregate:

aggr offline disaster_aggr

disaster_aggr is the name of the aggregate at the disaster site.


Note: An error message appears if the aggregate is already offline.

5. Re-create the mirrored aggregates by entering the following command for each aggregate thatwas split:

aggr mirror aggr_name -v disaster_aggr

aggr_name is the aggregate on the surviving site’s node.

disaster_aggr is the aggregate on the disaster site’s node.

The aggr_name aggregate rejoins the disaster_aggr aggregate to reestablish the MetroClusterconfiguration.

6. Verify that the mirrored aggregates have been re-created by entering the following command:

aggr status -r mir

The giveback operation only succeeds if the aggregates have been rejoined.

7. Enter the following command at the partner node:

cf giveback

The node at the disaster site reboots.

Example of rejoining aggregates

The following example shows the commands and status output when you rejoin aggregates toreestablish the MetroCluster configuration.

First, the aggregate status of the disaster site’s storage after reestablishing access to the partnernode at the surviving site is shown.

filer1> aggr status -rAggregate mir (online, normal) (zoned checksums) Plex /mir/plex5 (online, normal, active) RAID group /filer1/plex5/rg0 (normal)

RAID Disk Device HA SHELF BAY CHAN Used (MB/blks) Phys (MB/blks)--------- ------ ------------- ----- -------------- -------------parity 8a.2 8a 0 2 FC:B 34500/70656000 35003/71687368data 8a.8 8a 1 0 FC:B 34500/70656000 35003/71687368

Aggregate mir(1) (failed, out-of-date) (zoned checksums) Plex /mir(1)/plex1 (offline, normal, out-of-date) RAID group /mir(1)/plex1/rg0 (normal)


Plex /mir(1)/plex5 (offline, failed, out-of-date)

Next, the mirror is reestablished using the aggr mirror -v command.


Note: The node at the surviving site is called filer1; the node at the disaster site is calledfiler2.

filer1> aggr mirror mir -v mir(1)This will destroy the contents of mir(1). Are you sure? yMon Nov 18 15:36:59 GMT [filer1: raid.mirror.resync.snapcrtok:info]: mir: created mirror resynchronization snapshot mirror_resync.1118153658(filer2)Mon Nov 18 15:36:59 GMT [filer1: raid.rg.resync.start:notice]: /mir/plex6/rg0: start resynchronization (level 1)Mon Nov 18 15:36:59 GMT [filer1: raid.mirror.resync.start:notice]: /mir: start resynchronize to target /mir/plex6

After the aggregates rejoin, the synchronous mirrors of the MetroCluster configuration arereestablished.

filer1> aggr status -r mirAggregate mir (online, mirrored) (zoned checksums) Plex /mir/plex5 (online, normal, active) RAID group /mir/plex5/rg0 (normal)


Plex /mir/plex6 (online, normal, active) RAID group /mir/plex6/rg0 (normal)


Re-creating mirrored aggregates to return a MetroCluster to normal operation

Describes how to return the MetroCluster to normal operation by re-creating the MetroClustermirror.

Steps

1. Validate that you can access the remote storage by entering the following command:

aggr status -r

Note: A (level-0 resync in progress) message indicates that a plex cannot be rejoined.

2. Turn on the power to the node at the disaster site.

After the node at the disaster site boots up, it displays the following:

Waiting for Giveback...


3. If the aggregates at the disaster site are online, take them offline by entering the followingcommand for each aggregate that was split:

aggr offline disaster_aggr

disaster_aggr is the name of the aggregate at the disaster site.

Note: An error message appears if the aggregate is already offline.

4. Destroy every target plex that is in a level-0 resync state by entering the following command:

aggr destroy plex_name

For more information about the SyncMirror feature, see the Data ONTAP Data Protection OnlineBackup and Recovery Guide.

5. Re-create the mirrored aggregates by entering the following command for each aggregate thatwas split:

aggr mirror aggr_name

6. Enter the following command at the partner node:

cf giveback

The node at the disaster site reboots.

Example of re-creating a mirrored aggregate

The following example shows the commands and status output when re-creating aggregates toreestablish the MetroCluster configuration.

First, the aggregate status of the disaster site’s storage after reestablishing access to the partnerat the surviving site is shown.

filer1>aggr status -rAggregate mir1 (online, normal) (zoned checksums) Plex /mir1/plex0 (online, normal, active) RAID group /mir1/plex0/rg0 (normal)

RAID Disk Device HA SHELF BAY CHAN Used (MB/blks) Phys (MB/blks)--------- ------ ------------- ----- -------------- -------------parity 8a.3 8a 0 3 FC:B 34500/70656000 35003/71687368data 8a.4 8a 0 4 FC:B 34500/70656000 35003/71687368data 8a.6 8a 0 6 FC:B 34500/70656000 35003/71687368data 8a.5 8a 0 5 FC:B 34500/70656000 35003/71687368

Aggregate mir1(1) (failed, partial) (zoned checksums) Plex /mir1(1)/plex0 (offline, failed, inactive)

Plex /mir1(1)/plex6 (online, normal, resyncing) RAID group /mir1(1)/plex6/rg0 (level-0 resync in progress)

RAID Disk Device HA SHELF BAY CHAN Used (MB/blks) Phys (MB/blks)--------- ------ ------------- ----- -------------- -------------parity 6a.6 6a 0 6 FC:B 34500/70656000 35003/71687368


data 6a.2 6a 0 2 FC:B 34500/70656000 35003/71687368 data 6a.3 6a 0 3 FC:B 34500/70656000 35003/71687368 data 6a.5 6a 0 5 FC:B 34500/70656000 35003/71687368

The mir1(1)/plex6 plex shows that a level-0 resynchronization was in progress; therefore, anattempt to rejoin the plexes fails, as shown in the following output:

filer1> aggr mirror mir1 -v mir1(1)aggr mirror: Illegal mirror state for aggregate 'mir1(1)'

Because the mir1(1)/plex6 plex had a level-0 resynchronization in progress, the mir1(1)aggregate must be destroyed and the mir aggregate remirrored to reestablish a synchronousmirror, as shown in the following output:

filer1> aggr mirror mir1 -v mir1(1)aggr mirror: Illegal mirror state for aggregate 'mir1(1)'filer1> aggr destroy mir1(1)Are you sure you want to destroy this aggregate? yAggregate 'mir1(1)' destroyed.filer1> aggr mirror mir1Creation of a mirror plex with 4 disks has been initiated. The disks need to be zeroed before addition to the aggregate. The process has been initiated and you will be notified via the system log as disks are added.


Nondisruptive operations with active/activeconfigurations

By taking advantage of an active/active configuration's takeover and giveback operations, you canchange hardware components and perform software upgrades in your configuration withoutdisrupting access to system storage.

You can perform nondisruptive operations on a system by having its partner takeover the system'sstorage, performing maintenance, and then giving back the storage. Use the specific procedures asshown in the following table.

If you want to perform this tasknondisruptively...

See the...

Upgrade Data ONTAP Data ONTAP Upgrade Guide

Upgrade the controller Hardware System Upgrade Procedures

Replace a disk shelf Installation and Service Guide for your diskshelf

Replace a hardware FRU component FRU procedures for your platform

Nondisruptive operations with active/active configurations | 167

Controller failover and single-points-of-failure

A storage system has a variety of SPOFs that you can reduce by using an active/active configuration.In an active/active configuration, there are a number of failures that can cause a controller to failover.

Benefits of controller failover

You can use controller failover, a high-availability data service solution, to further increase theuptime of storage systems. It protects against controller failure by transferring the data service fromthe failed node to its partner node. Controller failover can also protect against other hardwarefailures, such as problems with network interface cards, Fibre Channel-Arbitrated Loops (FC-ALloops), SAS loops, and disk shelf modules. Controller failover is also an effective tool for reducingplanned downtime of one of the nodes.

Note: You might also see the term cluster failover; this is equivalent to the term controller failoverused in this document.

Single-point-of-failure definitionExplains what a single-point-of-failure is in the context of your storage system.

A single-point-of failure (SPOF) represents the failure of a single hardware component that can leadto loss of data access or potential loss of data. SPOF does not include multiple/rolling hardwareerrors, such as triple disk failure, dual disk shelf module failure, and so on.

All hardware components included with your storage system have demonstrated very good reliabilitywith low failure rates. If a hardware component fails, such as a controller or adapter, you can usecontroller failover to provide continuous data availability and preserve data integrity for clientapplications and users.

Controller failover and single-points-of-failure | 169

SPOF analysis for active/active configurationsEnables you to see which hardware components are SPOFs, and how controller failover caneliminate these SPOFs to improve data availability.

Hardwarecomponents

SPOF How controller failover eliminates SPOF

Stand-alone Active/activeconfiguration

Controller Yes No If a controller fails, the node automatically failsover to its partner node. The partner (takeover)node serves data for both of the nodes.

NVRAM Yes No If an NVRAM adapter fails, the nodeautomatically fails over to its partner node. Thepartner (takeover) node serves data for both ofthe nodes.

CPU fan Yes No If the CPU fan fails, the node gracefully shutsdown and automatically fails over to its partnernode. The partner (takeover) node serves datafor both of the nodes.

Multiple NICs withVIFs (virtualinterfaces)

No No If one of the networking links within a VIFfails, the networking traffic is automaticallysent over the remaining networking links onthe same node. No failover is needed in thissituation.

If all the NICs in a VIF fail, the nodeautomatically fails over to its partner node, iffailover is enabled for the VIF.

Single NIC Yes No If a NIC fails, the node automatically fails overto its partner node, if failover is enabled for theNIC.


Hardwarecomponents



FC-AL adapter or SASHBA

Yes No If an FC-AL adapter for the primary loop failsfor a configuration without multipath HA, thepartner node attempts a takeover at the time offailure. With multipath HA, no takeover isrequired.

If the FC-AL adapter for the secondary loopfails for a configuration without multipath HA,the failover capability is disabled, but bothnodes continue to serve data to their respectiveapplications and users, with no impact or delay.With multipath HA, failover capability is notaffected.

FC-AL or SAS cable(controller-to-shelf,shelf-to-shelf )

Yes No If an FC-AL loop or SAS stack breaks in aconfiguration that does not have multipath HA,the break could lead to a failover, depending onthe shelf type. The partnered nodes invoke thenegotiated failover feature to determine whichnode is best for serving data, based on the diskshelf count. When multipath HA is used, nofailover is required.

Disk shelf module Yes No If a disk shelf module fails in a configurationthat does not have multipath HA, the failurecould lead to a failover. The partnered nodesinvoke the negotiated failover feature todetermine which node is best for serving data,based on the disk shelf count. When multipathHA is used, there is no impact.

Disk drive No No If a disk fails, the node can reconstruct datafrom the RAID4 parity disk. No failover isneeded in this situation.


Hardwarecomponents



Power supply No No Both the controller and disk shelf have dualpower supplies. If one power supply fails, thesecond power supply automatically kicks in.No failover is needed in this situation. If bothpower supplies fail, the node automaticallyfails over to its partner node, which serves datafor both nodes.

Fan (controller or diskshelf)

No No Both the controller and disk shelf have multiplefans. If one fan fails, the second fanautomatically provides cooling. No failover isneeded in this situation. If both fans fail, thenode automatically fails over to its partnernode, which serves data for both nodes.

Cluster adapter N/A No If a cluster adapter fails, the failover capabilityis disabled but both nodes continue to servedata to their respective applications and users.

Cluster interconnectcable

N/A No The cluster interconnect adapter supports dualcluster interconnect cables. If one cable fails,the heartbeat and NVRAM data areautomatically sent over the second cable withno delay or interruption.

If both cables fail, the failover capability isdisabled but both nodes continue to serve datato their respective applications and users.


Failover event cause-and-effect tableFailover events cause a controller failover in active/active configurations. The configuration respondsdifferently depending on the event and the type of active/active configurations.

Cause-and-effect table for standard or mirrored active/active configurations

Event Does the eventtrigger failover?

Does the eventprevent a futurefailover fromoccurring, or afailover fromoccurringsuccessfully?

Does the event prevent a future failoverfrom occurring, or a failover fromoccurring successfully?

Single StorageSystem

Standard orMirrored

Single diskfailure

No No Yes Yes

Double diskfailure (2disks fail insame RAIDgroup)

Yes, unless youare usingSyncMirror orRAID-DP, thenno.

Maybe. If rootvolume hasdouble diskfailure or if themailbox disks areaffected, nofailover ispossible.

No, unless you areusing RAID-DP orSyncMirror, then yes.

No, unless you areusing RAID-DP orSyncMirror, thenyes.

Triple diskfailure (3disks fail insame RAIDgroup)

Maybe. IfSyncMirror isbeing used, therewon’t be atakeover;otherwise, yes.

Maybe. If rootvolume has tripledisk failure, nofailover ispossible.

No No

Single HBA(initiator)failure, LoopA

Maybe. IfSyncMirror ormultipath HA isin use, then no;otherwise, yes.

Maybe. If rootvolume hasdouble diskfailure, nofailover ispossible.

Yes, if multipath HAor SyncMirror isbeing used.

Yes, if multipath HAor SyncMirror isbeing used, or iffailover succeeds.






Standard orMirrored

Single HBA(initiator)failure, LoopB

No Yes, unless youare usingSyncMirror ormultipath HAand the mailboxdisks aren’taffected, then no.

Yes, if multipath HAor SyncMirror isbeing used.

Yes, if multipath HAor SyncMirror isbeing used, or iffailover succeeds.

Single HBAinitiatorfailure, (bothloops at thesame time)

Yes, unless thedata is mirroredon a different(up) loop ormultipath HA isin use, then notakeover needed.

Maybe. If thedata is mirroredor multipath HAis being used andthe mailbox disksare not affected,then no;otherwise, yes.

No, unless the data ismirrored or multipathHA is in use, thenyes.

No failover needed ifdata is mirrored ormultipath HA is inuse.

ESH2 or AT-FCX failure(Loop A)

Only if multidiskvolume failure oropen loopcondition occurs,and neitherSyncMirror normultipath HA isin use.


No Yes, if failoversucceeds.

ESH2 or AT-FCX failure(Loop B)

No Maybe. IfSyncMirror ormultipath HA isin use, then no;otherwise, yes.

Yes, if multipath HAor SyncMirror is inuse.

Yes






Standard orMirrored

IOM failure(Loop A)

Only if multidiskvolume failure oropen loopcondition occurs,and neitherSyncMirror normultipath HA isin use.



IOM failure(Loop B)

No Maybe. IfSyncMirror ormultipath HA isin use, then no;otherwise, yes.

Yes, if multipath HAor SyncMirror is inuse.

Yes

Shelf(backplane)failure

Only if multidiskvolume failure oropen loopcondition occurs,and data isn’tmirrored.

Maybe. If rootvolume hasdouble diskfailure or if themailboxes areaffected, nofailover ispossible.

Maybe. If data ismirrored, then yes;otherwise, no.


Shelf, singlepower failure

No No Yes Yes

Shelf, dualpower failure

Only if multidiskvolume failure oropen loopcondition occursand data isn’tmirrored.

Maybe. If rootvolume hasdouble diskfailure or if themailbox disks areaffected, nofailover ispossible.



Controller,single powerfailure

No No Yes Yes






Standard orMirrored

Controller,dual powerfailure

Yes Yes, until poweris restored.


Clusterinterconnectfailure (1port)

No No n/a Yes

Clusterinterconnectfailure (bothports)

No Yes n/a Yes

Ethernetinterfacefailure(primary, noVIF)

Yes, if set up todo so

No Yes Yes

Ethernetinterfacefailure(primary,VIF)


No Yes Yes

Ethernetinterfacefailure(secondary,VIF)


No Yes Yes

Ethernetinterfacefailure (VIF,all ports)


No Yes Yes

Tapeinterfacefailure

No No Yes Yes






Standard orMirrored

Heat exceedspermissibleamount

Yes No No No

Fan failures(disk shelvesor controller)

No No Yes Yes

Reboot No No Maybe. Depends oncause of reboot.

Maybe. Depends oncause of reboot.

Panic No No Maybe. Depends oncause of panic.

Maybe. Depends oncause of panic.

Cause-and-effect table for stretch and fabric-attached MetroClusters

Event Does theevent triggerfailover?

Does theeventprevent afuturefailover fromoccurring, ora failoverfromoccurringsuccessfully?

Is data still available on the affected volumeafter the event?

Stretch MetroCluster Fabric AttachedMetroCluster

Single diskfailure

No No Yes Yes

Double diskfailure (2 disksfail in sameRAID group)

No No Yes Yes, with no failovernecessary.

Triple diskfailure (3 disksfail in sameRAID group)

No No No Yes, with no failovernecessary.






Single HBA(initiator)failure, Loop A


Single HBA(initiator)failure, Loop B


Single HBAinitiator failure,(both loops atthe same time)

No No Yes, with no failovernecessary.

Yes, with no failovernecessary.

ESH2 or AT-FCX failure(Loop A)


ESH2 or AT-FCX failure(Loop B)

No No Yes Yes

Shelf(backplane)failure


Shelf, singlepower failure

No No Yes Yes

Shelf, dualpower failure


Controller,single powerfailure

No No Yes Yes

Controller,dual powerfailure

Yes Yes, untilpower isrestored.

Yes, if failoversucceeds.

Yes, if failoversucceeds.






Clusterinterconnectfailure (1 port)

No No Yes Yes

Clusterinterconnectfailure (bothports)

No No.

Failover ispossible.

Yes Yes

Ethernetinterfacefailure(primary, noVIF)

Yes, if set upto do so

No Yes Yes

Ethernetinterfacefailure(primary, VIF)


No Yes Yes

Ethernetinterfacefailure(secondary,VIF)


No Yes Yes

Ethernetinterfacefailure (VIF, allports)


No Yes Yes

Tape interfacefailure

No No Yes Yes

Heat exceedspermissibleamount

Yes No No No






Fan failures(disk shelves orcontroller)

No No Yes Yes

Reboot No No Maybe. Depends oncause of reboot.

Maybe. Depends oncause of reboot.

Panic No No Maybe. Depends oncause of panic.

Maybe. Depends oncause of panic.


Feature update record

Provides a record of the history of changes made to this guide. When a change is implemented, itapplies to the release in which it was implemented and all subsequent releases, unless otherwisespecified.

Feature updates Feature first implemented in Feature release date

• Updates for FAS920• Update for NVRAM5• Illustration updates

Data ONTAP 6.5.1 May 2004

• Updates for NVRAM5support in FAS900 seriesactive/active configurations,except for MetroCluster

• Failover event cause-and-effect table

• Declaration of Conformityupdate

• Addition of controllerfailover and single-point-of-failure analysis

Data ONTAP 7.0 November 2004

• FAS30xx information• Corrections were made to

the Upgrading an LRC toESH/ESH2/AT-FCXprocedure

Data ONTAP 7.0.1 April 2005

• Incorporation of the ClusterAdministration chapter fromthe Data ONTAP SystemAdministration Guide andthe Disaster ProtectionUsing MetroClusterappendix from the DataONTAP Data ProtectionOnline Backup andRecovery Guide.

Data ONTAP 7.1 June 2005

Feature update record | 181


• Updated MetroClusterinformation for FAS30xx

Data ONTAP 7.1 October 2005

• Updated modulereplacement information

• Fixed problem in Brocadeswitch configurationinformation

Data ONTAP 7.1 December 2005

• Updated and extendedactive/active configurationinformation

• Moved Brocade switchconfiguration to BrocadeSwitch Description Page.

• Moved from cluster toactive/active configuration

• Added information aboutMultipath Storage forActive/active configurations

Data ONTAP 7.1.1 June 2006

• Generalized standard andmirrored active/activeconfiguration cablinginstructions

• Updated standard andmirrored active/activeconfiguration cablinginstructions to includeFAS60xx

Data ONTAP 7.2 RC1 February 2006



• Changed name of documentfrom Cluster Installationand Administration Guideto Active/ActiveConfiguration Guide.

• Added FAS60xxinformation

• Updated and extendedactive/active configurationsconfiguration information

• Moved Brocade switchconfiguration to BrocadeSwitch Description Page.

• Moved from cluster toactive/active configuration

Data ONTAP 7.2 RC3 May 2006

• Added information aboutMultipath Storage forActive/activeconfigurations.

Data ONTAP 7.2.1 November 2006

• Added quad-port, 4-GbFibre Channel HBA, ESH4module, DS14mk4 FC diskshelf information

• Added information toexplain that automaticgiveback should not be usedin MetroClusters

• Updated Multipath Storageinformation

• Updated MetroClusterdisaster recoveryinformation

• Corrected failover andsingle-point-of-failure table

Data ONTAP 7.2.2 March 2007



• Added procedures forconfiguring fabric-attachedMetroClusters on systemsusing software-based diskmanagement

• Added procedure forunconfiguring an active/active pair and returning tostand-alone operation


• Added support for 504 disksin MetroClusters

• Added support for theFAS6040 and FAS6080systems

• Added support for thechange_fsid option

• Added procedure forremoving an active/activeconfiguration

Data ONTAP 7.2.4 November 2007

• Added support forFAS2020/FAS2050andFAS3140/FAS3170 systems

• Added procedures for thehardware-assisted takeoverfeature

• Added information aboutdisk shelves on V-SeriesActive/active configurations




• Added V-Series content• Added references to the

DS4243 disk shelfdocumentation

• Added support for 672 disksin MetroClusters

• Added MetroClustersupport for the Brocade 300and 5100 switches

• Add support forMetroCluster nodes onseparate subnetworks

Data ONTAP 7.3.2 September 2009

• Added support for 32xxsystems

• Added support for the 8-Gbps FC-VI adapter onMetroCluster configurations

Data ONTAP 7.3.5 October 2010

• Added more SAS disk shelfinformation and referencesto SAS disk shelfdocumentation.

• Updated multipath HA as arequirement.

• Removed procedures fornon-multipath HA cabling.

• Replaced the termMultipath Storage withmultipath HA forconsistency with otherdocumentation.

Data ONTAP 8.0.2

Data ONTAP 7.3.5.1

May 2011


Abbreviations

A list of abbreviations and their spelled-out forms are included here for your reference.

A

ABE (Access-Based Enumeration)

ACE (Access Control Entry)

ACL (access control list)

ACP (Alternate Control Path)

AD (Active Directory)

ALPA (arbitrated loop physical address)

ALUA (Asymmetric Logical Unit Access)

AMS (Account Migrator Service)

API (Application Program Interface)

ARP (Address Resolution Protocol)

ASCII (American Standard Code for Information Interchange)

ASP (Active Server Page)

ATA (Advanced Technology Attachment)

B

BCO (Business Continuance Option)

BIOS (Basic Input Output System

BCS (block checksum type )

BLI (block-level incremental)

BMC (Baseboard Management Controller)

Abbreviations | 187

C

CD-ROM (compact disc read-only memory)

CDDI (Copper Distributed Data Interface)

CDN (content delivery network)

CFE (Common Firmware Environment)

CFO (controller failover)

CGI (Common Gateway Interface)

CHA (channel adapter)

CHAP (Challenge Handshake Authentication Protocol)

CHIP (Client-Host Interface Processor)

CIDR (Classless Inter-Domain Routing)

CIFS (Common Internet File System)

CIM (Common Information Model)

CLI (command-line interface)

CP (consistency point)

CPU (central processing unit)

CRC (cyclic redundancy check)

CSP (communication service provider)


D

DAFS (Direct Access File System)

DBBC (database consistency checker)

DCE (Distributed Computing Environment)

DDS (Decru Data Decryption Software)

dedupe (deduplication)

DES (Data Encryption Standard)

DFS (Distributed File System)

DHA (Decru Host Authentication)

DHCP (Dynamic Host Configuration Protocol)

DIMM (dual-inline memory module)

DITA (Darwin Information Typing Architecture)

DLL (Dynamic Link Library)

DMA (direct memory access)

DMTD (Distributed Management Task Force)

DNS (Domain Name System)

DOS (Disk Operating System)

DPG (Data Protection Guide)

DTE (Data Terminal Equipment)

Abbreviations | 189

E

ECC (Elliptic Curve Cryptography) or (EMC Control Center)

ECDN (enterprise content delivery network)

ECN (Engineering Change Notification)

EEPROM (electrically erasable programmable read-only memory)

EFB (environmental fault bus)

EFS (Encrypted File System)

EGA (Enterprise Grid Alliance)

EISA (Extended Infrastructure Support Architecture)

ELAN (Emulated LAN)

EMU environmental monitoring unit)

ESH (embedded switching hub)

F

FAQs (frequently asked questions)

FAS (fabric-attached storage)

FC (Fibre Channel)

FC-AL (Fibre Channel-Arbitrated Loop)

FC SAN (Fibre Channel storage area network)

FC Tape SAN (Fibre Channel Tape storage area network)

FC-VI (virtual interface over Fibre Channel)

FCP (Fibre Channel Protocol)

FDDI (Fiber Distributed Data Interface)

FQDN (fully qualified domain name)

FRS (File Replication Service)

FSID (file system ID)

FSRM (File Storage Resource Manager)

FTP (File Transfer Protocol)


G

GbE (Gigabit Ethernet)

GID (group identification number)

GMT (Greenwich Mean Time)

GPO (Group Policy Object)

GUI (graphical user interface)

GUID (globally unique identifier)

H

HA (high availability)

HBA (host bus adapter)

HDM (Hitachi Device Manager Server)

HP (Hewlett-Packard Company)

HTML (hypertext markup language)

HTTP (Hypertext Transfer Protocol)

Abbreviations | 191

I

IB (InfiniBand)

IBM (International Business Machines Corporation)

ICAP (Internet Content Adaptation Protocol)

ICP (Internet Cache Protocol)

ID (identification)

IDL (Interface Definition Language)

ILM (information lifecycle management)

IMS (If-Modified-Since)

I/O (input/output)

IP (Internet Protocol)

IP SAN (Internet Protocol storage area network)

IQN (iSCSI Qualified Name)

iSCSI (Internet Small Computer System Interface)

ISL (Inter-Switch Link)

iSNS (Internet Storage Name Service)

ISP (Internet storage provider)

J

JBOD (just a bunch of disks)

JPEG (Joint Photographic Experts Group)

K

KB (Knowledge Base)

Kbps (kilobits per second)

KDC (Kerberos Distribution Center)


L

LAN (local area network)

LBA (Logical Block Access)

LCD (liquid crystal display)

LDAP (Lightweight Directory Access Protocol)

LDEV (logical device)

LED (light emitting diode)

LFS (log-structured file system)

LKM (Lifetime Key Management)

LPAR (system logical partition)

LRC (Loop Resiliency Circuit)

LREP (logical replication tool utility)

LUN (logical unit number)

LUSE (Logical Unit Size Expansion)

LVM (Logical Volume Manager)

Abbreviations | 193

M

MAC (Media Access Control)

Mbps (megabits per second)

MCS (multiple connections per session)

MD5 (Message Digest 5)

MDG (managed disk group)

MDisk (managed disk)

MIB (Management Information Base)

MIME (Multipurpose Internet Mail Extension)

MMC (Microsoft Management Console)

MMS (Microsoft Media Streaming)

MPEG (Moving Picture Experts Group)

MPIO (multipath network input/output)

MRTG (Multi-Router Traffic Grapher)

MSCS (Microsoft Cluster Service

MSDE (Microsoft SQL Server Desktop Engine)

MTU (Maximum Transmission Unit)


N

NAS (network-attached storage)

NDMP (Network Data Management Protocol)

NFS (Network File System)

NHT (NetApp Health Trigger)

NIC (network interface card)

NMC (Network Management Console)

NMS (network management station)

NNTP (Network News Transport Protocol)

NTFS (New Technology File System)

NTLM (NetLanMan)

NTP (Network Time Protocol)

NVMEM (nonvolatile memory management)

NVRAM (nonvolatile random-access memory)

O

OFM (Open File Manager)

OFW (Open Firmware)

OLAP (Online Analytical Processing)

OS/2 (Operating System 2)

OSMS (Open Systems Management Software)

OSSV (Open Systems SnapVault)

Abbreviations | 195

P

PC (personal computer)

PCB (printed circuit board)

PCI (Peripheral Component Interconnect)

pcnfsd (storage daemon)

(PC)NFS (Personal Computer Network File System)

PDU (protocol data unit)

PKI (Public Key Infrastructure)

POP (Post Office Protocol)

POST (power-on self-test)

PPN (physical path name)

PROM (programmable read-only memory)

PSU power supply unit)

PVC (permanent virtual circuit)

Q

QoS (Quality of Service)

QSM (Qtree SnapMirror)


R

RAD (report archive directory)

RADIUS (Remote Authentication Dial-In Service)

RAID (redundant array of independent disks)

RAID-DP (redundant array of independent disks, double-parity)

RAM (random access memory)

RARP (Reverse Address Resolution Protocol)

RBAC (role-based access control)

RDB (replicated database)

RDMA (Remote Direct Memory Access)

RIP (Routing Information Protocol)

RISC (Reduced Instruction Set Computer)

RLM (Remote LAN Module)

RMC (remote management controller)

ROM (read-only memory)

RPM (revolutions per minute)

rsh (Remote Shell)

RTCP (Real-time Transport Control Protocol)

RTP (Real-time Transport Protocol)

RTSP (Real Time Streaming Protocol)

Abbreviations | 197

S

SACL (system access control list)

SAN (storage area network)

SAS (storage area network attached storage) or (serial-attached SCSI)

SATA (serial advanced technology attachment)

SCSI (Small Computer System Interface)

SFO (storage failover)

SFSR (Single File SnapRestore operation)

SID (Secure ID)

SIMM (single inline memory module)

SLB (Server Load Balancer)

SLP (Service Location Protocol)

SNMP (Simple Network Management Protocol)

SNTP (Simple Network Time Protocol)

SP (Storage Processor)

SPN (service principal name)

SPOF (single point of failure)

SQL (Structured Query Language)

SRM (Storage Resource Management)

SSD (solid state disk

SSH (Secure Shell)

SSL (Secure Sockets Layer)

STP (shielded twisted pair)

SVC (switched virtual circuit)


T

TapeSAN (tape storage area network)

TCO (total cost of ownership)

TCP (Transmission Control Protocol)

TCP/IP (Transmission Control Protocol/Internet Protocol)

TOE (TCP offload engine)

TP (twisted pair)

TSM (Tivoli Storage Manager)

TTL (Time To Live)

U

UDP (User Datagram Protocol)

UI (user interface)

UID (user identification number)

Ultra ATA (Ultra Advanced Technology Attachment)

UNC (Uniform Naming Convention)

UPS (uninterruptible power supply)

URI (universal resource identifier)

URL (uniform resource locator)

USP (Universal Storage Platform)

UTC (Universal Coordinated Time)

UTP (unshielded twisted pair)

UUID (universal unique identifier)

UWN (unique world wide number)

Abbreviations | 199

V

VCI (virtual channel identifier)

VCMDB (Volume Configuration Management Database)

VDI (Virtual Device Interface)

VDisk (virtual disk)

VDS (Virtual Disk Service)

VFM (Virtual File Manager)

VFS (virtual file system)

VI (virtual interface)

vif (virtual interface)

VIRD (Virtual Router ID)

VLAN (virtual local area network)

VLD (virtual local disk)

VOD (video on demand)

VOIP (voice over IP)

VRML (Virtual Reality Modeling Language)

VTL (Virtual Tape Library)

W

WAFL (Write Anywhere File Layout)

WAN (wide area network)

WBEM (Web-Based Enterprise Management)

WHQL (Windows Hardware Quality Lab)

WINS (Windows Internet Name Service)

WORM (write once, read many)

WWN (worldwide name)

WWNN (worldwide node name)

WWPN (worldwide port name)

www (worldwide web)


Z

ZCS (zoned checksum)

Abbreviations | 201

Glossary

Storage terms

array LUN The storage that third-party storage arrays provide to storage systems runningData ONTAP software. One array LUN is the equivalent of one disk on anative disk shelf.

LUN (logical unitnumber)

A logical unit of storage identified by a number.

ESH (EmbeddedSwitching Hub)disk shelf module

A component that provides a means of managing an FC-AL loop in anintelligent manner, such that a single drive failure does not take down the loop.It also contains the enclosure services processor, which communicates theenvironmental data of the disk shelf.

ESH2 disk shelfmodule

A second-generation ESH module.

ESH4 disk shelfmodule

A third-generation ESH module.

multipath HA In an active/active configuration, a configuration in which each controller hasmultiple ways to connect to a disk drive. Multipath HA cabling is the mostresilient and preferred configuration for active/active configurations. This isbecause it takes full advantage of the resiliency capability of the disk shelves,which means that the node continues to have access to disk drives in the eventof cable, HBA, or shelf module failure. A single failure of a cable, HBA, ormodule does not result in a controller failover.

native disk A disk that is sold as local storage for storage systems that run Data ONTAPsoftware.

native disk shelf A disk shelf that is sold as local storage for storage systems that run DataONTAP software.

SAS stack Also referred to as stack. A group of one or more SAS disk shelves connected(daisy-chained) together and connected to the controller through the top diskshelf in the stack and the bottom disk shelf in the stack (as needed). Themaximum number of disk shelves in a stack of disk shelves and the number ofdisk shelf stacks supported in a configuration are dependent on the type ofstorage system.

storage controller The component of a storage system that runs the Data ONTAP operatingsystem and controls its disk subsystem. Storage controllers are also sometimes

Glossary | 203

called controllers, storage appliances, appliances, storage engines, heads, CPUmodules, or controller modules.

storage system The hardware device running Data ONTAP that receives data from and sendsdata to native disk shelves, third-party storage, or both. Storage systems thatrun Data ONTAP are sometimes referred to as filers, appliances, storageappliances, V-Series systems, or systems.

third-partystorage

The back-end storage arrays, such as IBM, Hitachi Data Systems, and HP, thatprovide storage for storage systems running Data ONTAP.

Cluster and high-availability terms

active/activeconfiguration

• In the Data ONTAP 7.2 and 7.3 release families, a pair of storage systems orV-Series systems (sometimes called nodes) configured to serve data for eachother if one of the two systems stops functioning. Also sometimes referred toas active/active pairs.

• In the Data ONTAP 8.x release family, this functionality is referred to as ahigh-availability (HA) configuration or an HA pair.

• In the Data ONTAP 7.1 release family and earlier releases, this functionalityis referred to as a cluster.

MetroCluster A type of HA pair that provides the capability to force a takeover when an entirenode is destroyed or unavailable. Not all platforms, switches, storage subsystems,or Data ONTAP versions are supported in MetroCluster configurations.


Copyright information

Copyright © 1994–2011 NetApp, Inc. All rights reserved. Printed in the U.S.A.

No part of this document covered by copyright may be reproduced in any form or by any means—graphic, electronic, or mechanical, including photocopying, recording, taping, or storage in anelectronic retrieval system—without prior written permission of the copyright owner.

Software derived from copyrighted NetApp material is subject to the following license anddisclaimer:

THIS SOFTWARE IS PROVIDED BY NETAPP "AS IS" AND WITHOUT ANY EXPRESS ORIMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIEDWARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE,WHICH ARE HEREBY DISCLAIMED. IN NO EVENT SHALL NETAPP BE LIABLE FOR ANYDIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIALDAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTEGOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESSINTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHERIN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OROTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IFADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

NetApp reserves the right to change any products described herein at any time, and without notice.NetApp assumes no responsibility or liability arising from the use of products described herein,except as expressly agreed to in writing by NetApp. The use or purchase of this product does notconvey a license under any patent rights, trademark rights, or any other intellectual property rights ofNetApp.

The product described in this manual may be protected by one or more U.S.A. patents, foreignpatents, or pending applications.

RESTRICTED RIGHTS LEGEND: Use, duplication, or disclosure by the government is subject torestrictions as set forth in subparagraph (c)(1)(ii) of the Rights in Technical Data and ComputerSoftware clause at DFARS 252.277-7103 (October 1988) and FAR 52-227-19 (June 1987).

Copyright information | 205

Trademark information

NetApp, the NetApp logo, Network Appliance, the Network Appliance logo, Akorri,ApplianceWatch, ASUP, AutoSupport, BalancePoint, BalancePoint Predictor, Bycast, CampaignExpress, ComplianceClock, Cryptainer, CryptoShred, Data ONTAP, DataFabric, DataFort, Decru,Decru DataFort, FAServer, FilerView, FlexCache, FlexClone, FlexPod, FlexScale, FlexShare,FlexSuite, FlexVol, FPolicy, GetSuccessful, gFiler, Go further, faster, Imagine Virtually Anything,Lifetime Key Management, LockVault, Manage ONTAP, MetroCluster, MultiStore, NearStore,NetCache, NOW (NetApp on the Web), Onaro, OnCommand, ONTAPI, OpenKey, RAID-DP,ReplicatorX, SANscreen, SecureAdmin, SecureShare, Select, Service Builder, Shadow Tape,Simulate ONTAP, SnapCopy, SnapDirector, SnapDrive, SnapFilter, SnapLock, SnapManager,SnapMigrator, SnapMirror, SnapMover, SnapProtect, SnapRestore, Snapshot, SnapSuite,SnapValidator, SnapVault, StorageGRID, StoreVault, the StoreVault logo, SyncMirror, Tech OnTap,The evolution of storage, Topio, vFiler, VFM, Virtual File Manager, VPolicy, WAFL, and Web Filerare trademarks or registered trademarks of NetApp, Inc. in the United States, other countries, or both.

IBM, the IBM logo, and ibm.com are trademarks or registered trademarks of International BusinessMachines Corporation in the United States, other countries, or both. A complete and current list ofother IBM trademarks is available on the Web at www.ibm.com/legal/copytrade.shtml.

Apple is a registered trademark and QuickTime is a trademark of Apple, Inc. in the U.S.A. and/orother countries. Microsoft is a registered trademark and Windows Media is a trademark of MicrosoftCorporation in the U.S.A. and/or other countries. RealAudio, RealNetworks, RealPlayer,RealSystem, RealText, and RealVideo are registered trademarks and RealMedia, RealProxy, andSureStream are trademarks of RealNetworks, Inc. in the U.S.A. and/or other countries.

All other brands or products are trademarks or registered trademarks of their respective holders andshould be treated as such.

NetApp, Inc. is a licensee of the CompactFlash and CF Logo trademarks.

NetApp, Inc. NetCache is certified RealSystem compatible.

Trademark information | 207

http://www.ibm.com/legal/copytrade.shtml

How to send your comments

You can help us to improve the quality of our documentation by sending us your feedback.

Your feedback is important in helping us to provide the most accurate and high-quality information.If you have suggestions for improving this document, send us your comments by e-mail to [email protected]. To help us direct your comments to the correct division, include in thesubject line the name of your product and the applicable operating system. For example, FAS6070—Data ONTAP 7.3, or Host Utilities—Solaris, or Operations Manager 3.8—Windows.

You can also contact us in the following ways:

• NetApp, Inc., 495 East Java Drive, Sunnyvale, CA 94089• Telephone: +1 (408) 822-6000• Fax: +1 (408) 822-4501• Support Telephone: +1 (888) 4-NETAPP

How to send your comments | 209

mailto:[email protected]

IndexA

active/active configurationsbenefits of 9changing nodes to stand-alone 91–93, 95, 96characteristics of 10converting to MetroCluster 63definition of 9restrictions 17setup requirements for 17types of

compared 14fabric-attached MetroClusters 28installed in equipment racks 36installed in system cabinets 36mirrored 21standard 15stretch MetroClusters 23

adaptersquad-port Fibre Channel HBA 41, 48

aggregatesrecreating mirrored after disaster 163rejoining after disaster 161

automatic giveback 144

B

bring upconfiguring interfaces for 109manually setting options for 103

Brocade switch configurationswitch bank rules 75virtual channel rules 75

C

cable 38, 61cabling

Channel Afor mirrored active/active configuration 50for standard active/active configuration 42

Channel Bfor mirrored active/active configuration 52for standard active/active configuration 44

cluster interconnect for fabric-attachedMetroClusters

with software-based disk ownership 85cluster interconnect for standard active/active

configuration 46, 56cluster interconnect for standard active/active

configuration, 32xx systems 47, 57cross-cabled cluster interconnect 47, 57cross-cabled HA interconnect 46, 56error message, cross-cabled cluster interconnect 46,

47, 56, 57fabric-attached MetroClusters 71FC-VI adapter for fabric-attached MetroClusters

with software-based disk ownership 80, 85local controller in fabric-attached MetroCluster

with software-based disk ownership 76local disk shelves in fabric-attached MetroCluster

with software-based disk ownership 78preparing equipment racks for 39preparing system cabinets for 40remote controller in fabric-attached MetroCluster

with software-based disk ownership 82remote disk shelves in fabric-attached MetroCluster

with software-based disk ownership 84requirements 38, 61stretch MetroClusters 67

cf forcegiveback command 142cf giveback command 140cf-config-check.cgi utility 119cf.giveback.auto.cifs.terminate.minutes options 143cf.giveback.auto.enable option 144cf.giveback.auto.terminate.bigjobs option 143cf.giveback.check.partner option 143cf.takeover.on_network_interface_failure option 133cf.takeover.on_network_interface_failure.policy option

133cf.takeover.use_mcrc_file 118change_fsid option 104Channel A

cabling 42, 50defined 22

Channel Bcabling 44, 52

checking configuration through a utility 119CIFS clients and giveback delay 143CIFS sessions terminated on takeover 122cluster interconnect, cabling 80, 85command exceptions for emulated nodes 137, 138

Index | 211

commandscf (enables and disables takeover) 128cf forcesgiveback (forces giveback) 142cf forcetakeover -d (forces takeover) 158cf forcetakeover (forces takeover 131cf giveback (enables giveback) 119cf giveback (initiates giveback) 140cf partner (displays partner's name) 127cf status (displays status) 123, 134cf takeover (initiates takeover 131cf takeover (initiates takeover) 119halt (halts system without takeover) 128license add (license cluster) 102partner (accesses emulated node) 135storage show disk -p (displays paths) 151sysconfig 127takeover (description of all takeover commands)

131comparison of types of active/active configurations 14configuration speeds

changing stretch MetroCluster default 68resetting stretch MetroCluster default 70

configuration variationsfabric-attached MetroCluster configurations 34mirrored active/active configurations 23standard active/active configurations 18stretch MetroClusters 28

configurationsreestablishing MetroCluster configuration 161testing 119

configuringdedicated and standby interfaces 113shared interfaces 113

controller failoverbenefits 169

controller-to-switch cabling, fabric-attachedMetroClusters 76, 82

DData ONTAP

in a standard active/active configurations 15in fabric-attached MetroCluster configurations 31in stretch MetroCluster configurations 26

dedicated interfacesconfiguring using ifconfig 113configuring using setup 101described 110diagram 112

delay, specifying before takeover 132disabling takeover (cf) 128

disastersdetermining whether one occurred 155recognizing 155recovery from

forcing takeover 158manually fencing off the disaster site node 157reestablishing MetroCluster configuration 161restricting access to the failed node 157steps 156using MetroCluster 155when not to perform 155

disk information, displaying 127disk paths, verifying in a fabric-attached MetroCluster

with software-based disk ownership 88disk shelf pool assignments, fabric-attached

MetroClusters 87disk shelves

about modules for 149adding to an active/active configuration with

multipath HA 147comparison 133hot swapping modules in 153

disk-shelf-to-switch cabling, fabric-attachedMetroClusters 78, 84

documentation, required 36, 59dual-chassis HA configurations

diagram 10interconnect 11

E

e0M management interface 112eliminating single-point-of-failure (SPOF) 170emulated LANs

considerations for 113emulated node

accessing from the takeover node 135accessing remotely 136backing up 139commands that are unavailable in 137, 138description of 135dumps and restores 139managing 135

enabling takeover (cf) 128equipment racks

installation in 36preparation of 39

events triggering failover 173, 177


F

fabric-attached MetroCluster configurationassigning disk pools 87behavior of Data ONTAP with 31local node

cabling controller to switchwith software-based disk ownership 76

cabling disk shelves to switchwith software-based disk ownership 78

remote nodecabling controller to switch

with software-based disk ownership 82cabling disk shelves to switch

with software-based disk ownership 84verifying disk paths

with software-based disk ownership 88fabric-attached MetroClusters

about 28advantages of 29Brocade switch configuration 75cabling 71, 76, 78, 82, 84illustration of 71limitations 33planning worksheet 73restrictions 31–33setup requirements for 31–33upgrading from hardware-based to software-based

disk ownership 65variations 34

fabric-attached MetroClusters configurationcabling cluster interconnect for

cabling FC-VI adapter forwith software-based disk ownership 80,

85with software-based disk ownership 80, 85

failovercause-and-effect table 173, 177determining status (cf status) 134

failures that trigger failover 173, 177FC-VI adapter, cabling 80, 85fencing, manual 157Fibre Channel ports

identifying for active/active configuration 41, 48Fibre Channel switches 61forcing

giveback 142takeover 131

G

givebackautomatic 144automatically terminating long-running processes

143delay time for CIFS clients 143description of 122managing 140normal 140performing a 140shortening 143testing 119troubleshooting 144

H

HA configuration checker 119HA configuration variations 10HA interconnect

cabling 46, 56cabling, 32xx dual-chassis HA configurations 47,

57single-chassis and dual-chassis HA configurations

11HA state 11, 105ha-config modify command 11, 105ha-config show command 11, 105halting system without takeover 128hardware

active/active components described 16components described 16single-point-of-failure 169upgrading nondisruptively 167

hardware assisted takeover 106, 107hardware-based disk ownership 65

I

immediate takeover, enabling or disabling 128installation

equipment rack 36system cabinet 36

interface configurationsdedicated 110shared 110standby 110

interfacesconfiguration for takeover 112configuring dedicated 101

Index | 213

configuring for automatic takeover 133configuring shared 100configuring standby 101dedicated, diagram 112shared, diagram 111standby, diagram 112types and configurations 109, 110, 112

IPv6 considerations 109, 110

L

licensesenabling cluster 102required 102

lun commands, lun online 159LUNs, bringing online 159

M

mailbox disks 10managing in normal mode 123manual fencing 157MetroCluster

resetting the default speed of stretch 70MetroClusters

changing the default speed of stretch 68converting to from a standard or mirrored active/

active configuration 63disaster recovery using 155LUNs and 159reestablishing configuration after disaster 161software-based disk ownership and 63

mirrored active/active configurationcabling Channel A 50cabling Channel B 52

mirrored active/active configurationsabout 21advantages of 22restrictions 22setup requirements for 22variations 23

modules, disk shelfabout 149best practices for changing types 150hot-swapping 153restrictions for changing types 149testing 150

multipath HAadvantages of 20best practices 20, 21

connection types used by 19description of 18requirements 20, 21

N

network interfacesconfiguration for takeover 112configuring for takeover 113emulated LAN considerations 113types and configurations 109, 110, 112

nondisruptive upgrades, hardware 167normal mode

managing in 123NVRAM adapter 38, 61

O

options, setting 103

P

parameterschange_fsid 104required to be identical between nodes 104setting 103

partner command 135partner name, displaying (cf partner) 127planning worksheet for fabric-attached MetroClusters 73pool assignments, fabric-attached MetroClusters 87port list

creating for mirrored active/active configurations49

portsidentifying which ones to use 41, 48

preparing equipment racks 39primary connections, in multipath HA 19

R

redundant connections, in multipath HA 19reestablishing MetroCluster configuration 161removing an active/active configuration 91requirements

adapters 61documentation 36, 59equipment 38, 61Fibre Channel switches 61


for upgrading to a fabric-attached MetroClusterusing software-based disk ownership 65

hot-swapping a disk shelf module 153multipath HA 20, 21NVRAM adapter 61SFP modules 61tools 38, 60

restrictionsfabric-attached MetroCluster 31–33in active/active configurations 17in mirrored active/active configurations 22in stretch MetroClusters 27

rsh, using to access node after takeover 122

S

setting options and parameters 103setup, running on active/active configurations 99SFP modules 38, 61shared interfaces

configuring using ifconfig 113configuring using setup 100described 110diagram 111

single-chassis HA configurationsdiagram 10interconnect 11

single-point-of-failure (SPOF), eliminating 170single-point-of-failure, definition of 169SNMP protocol and takeover mode 134software-based disk management 87software-based disk ownership 63, 65, 76, 78, 82, 84SPOF (single-point-of-failure) 169stand-alone operation

changing a node in an active/active configuration to91–93, 95, 96

standard active/active configurationcabling Channel A 42cabling Channel B 44cabling cluster interconnect for 46, 56cabling cluster interconnect for, 32xx systems 47,

57contents of 15variations 18

standard active/active configurationsbehavior of Data ONTAP with 15

standby connections, in multipath HA 19standby interfaces

configuring using ifconfig 113configuring using setup 101

described 110diagram 112

status messages, descriptions of 126status, monitoring active/active pair 123stretch MetroCluster

resetting default speed 70stretch MetroClusters

about 23advantages of 24behavior of Data ONTAP with 26cabling 67changing the default speed of 68connections required 25illustration of 24, 25on dual-controller systems 26restrictions 27variations 28

switch configuration, for fabric-attached MetroClusters75

system cabinetsinstallation in 36preparing for cabling 40

T

takeoverCIFS sessions and 122configuring VIFs for automatic 133configuring when it occurs 129configuring with dedicated and hot standby

interfaces 112determining why one occurred 134disabling 128disabling immediate 128enabling 128enabling immediate 128forcing 131forcing for disaster recovery 158hardware assisted 106, 107reasons for 129rsh access after 122SNMP settings and 134specifying delay before 132statistics 134telnet access after 122testing 119troubleshooting 144using /etc/mcrc file at takeover 118what happens after 122what happens during 122

Index | 215

when it occurs 121takeover mode

managing in 134statistics in 134

telnet, using to access node after takeover 122tools, required 38, 60

U

unconfiguring an active/active pair 91upgrading

hardware, nondisruptively 167UPS

using with active/active configurations 57using with MetroCluster configurations 88

utility, cf-config-check.cgi 119

V

VIFsconfiguring for automatic takeover 133using to reduce SPOF 99


netapp

Documents

lifetime key management

softwarebased disk ownership

based disk ownership

series support matrix

metrocluster compatibility matrix

common gateway interface

disk shelf models

disk shelf count